Compare commits

..

1 Commits

Author SHA1 Message Date
John Crowley
d4326bfaf4 Initial implementation of GCS provider. 2025-04-22 13:46:57 -05:00
42 changed files with 1357 additions and 668 deletions

View File

@@ -19,7 +19,7 @@
!pageserver/
!pgxn/
!proxy/
!endpoint_storage/
!object_storage/
!storage_scrubber/
!safekeeper/
!storage_broker/

View File

@@ -133,7 +133,6 @@ runs:
fi
PERF_REPORT_DIR="$(realpath test_runner/perf-report-local)"
echo "PERF_REPORT_DIR=${PERF_REPORT_DIR}" >> ${GITHUB_ENV}
rm -rf $PERF_REPORT_DIR
TEST_SELECTION="test_runner/${{ inputs.test_selection }}"
@@ -210,12 +209,11 @@ runs:
--verbose \
-rA $TEST_SELECTION $EXTRA_PARAMS
- name: Upload performance report
if: ${{ !cancelled() && inputs.save_perf_report == 'true' }}
shell: bash -euxo pipefail {0}
run: |
export REPORT_FROM="${PERF_REPORT_DIR}"
scripts/generate_and_push_perf_report.sh
if [[ "${{ inputs.save_perf_report }}" == "true" ]]; then
export REPORT_FROM="$PERF_REPORT_DIR"
export REPORT_TO="$PLATFORM"
scripts/generate_and_push_perf_report.sh
fi
- name: Upload compatibility snapshot
# Note, that we use `github.base_ref` which is a target branch for a PR

140
Cargo.lock generated
View File

@@ -2037,33 +2037,6 @@ dependencies = [
"zeroize",
]
[[package]]
name = "endpoint_storage"
version = "0.0.1"
dependencies = [
"anyhow",
"axum",
"axum-extra",
"camino",
"camino-tempfile",
"futures",
"http-body-util",
"itertools 0.10.5",
"jsonwebtoken",
"prometheus",
"rand 0.8.5",
"remote_storage",
"serde",
"serde_json",
"test-log",
"tokio",
"tokio-util",
"tower 0.5.2",
"tracing",
"utils",
"workspace_hack",
]
[[package]]
name = "enum-map"
version = "2.5.0"
@@ -2451,6 +2424,33 @@ dependencies = [
"slab",
]
[[package]]
name = "gcp_auth"
version = "0.12.3"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "dbf67f30198e045a039264c01fb44659ce82402d7771c50938beb41a5ac87733"
dependencies = [
"async-trait",
"base64 0.22.1",
"bytes",
"chrono",
"home",
"http 1.1.0",
"http-body-util",
"hyper 1.4.1",
"hyper-rustls 0.27.5",
"hyper-util",
"ring",
"rustls-pemfile 2.1.1",
"serde",
"serde_json",
"thiserror 1.0.69",
"tokio",
"tracing",
"tracing-futures",
"url",
]
[[package]]
name = "gen_ops"
version = "0.4.0"
@@ -2749,6 +2749,15 @@ dependencies = [
"digest",
]
[[package]]
name = "home"
version = "0.5.11"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "589533453244b0995c858700322199b2becb13b627df2851f64a2775d024abcf"
dependencies = [
"windows-sys 0.59.0",
]
[[package]]
name = "hostname"
version = "0.4.0"
@@ -2978,6 +2987,24 @@ dependencies = [
"tower-service",
]
[[package]]
name = "hyper-rustls"
version = "0.27.5"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "2d191583f3da1305256f22463b9bb0471acad48a4e534a5218b9963e9c1f59b2"
dependencies = [
"futures-util",
"http 1.1.0",
"hyper 1.4.1",
"hyper-util",
"rustls 0.23.18",
"rustls-native-certs 0.8.0",
"rustls-pki-types",
"tokio",
"tokio-rustls 0.26.0",
"tower-service",
]
[[package]]
name = "hyper-timeout"
version = "0.5.1"
@@ -3733,6 +3760,16 @@ version = "0.3.17"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "6877bb514081ee2a7ff5ef9de3281f14a4dd4bceac4c09388074a6b5df8a139a"
[[package]]
name = "mime_guess"
version = "2.0.5"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "f7c44f8e672c00fe5308fa235f821cb4198414e1c77935c1ab6948d3fd78550e"
dependencies = [
"mime",
"unicase",
]
[[package]]
name = "minimal-lexical"
version = "0.2.1"
@@ -4025,6 +4062,33 @@ dependencies = [
"memchr",
]
[[package]]
name = "object_storage"
version = "0.0.1"
dependencies = [
"anyhow",
"axum",
"axum-extra",
"camino",
"camino-tempfile",
"futures",
"http-body-util",
"itertools 0.10.5",
"jsonwebtoken",
"prometheus",
"rand 0.8.5",
"remote_storage",
"serde",
"serde_json",
"test-log",
"tokio",
"tokio-util",
"tower 0.5.2",
"tracing",
"utils",
"workspace_hack",
]
[[package]]
name = "once_cell"
version = "1.20.2"
@@ -5522,8 +5586,11 @@ dependencies = [
"bytes",
"camino",
"camino-tempfile",
"chrono",
"futures",
"futures-util",
"gcp_auth",
"http 1.1.0",
"http-body-util",
"http-types",
"humantime-serde",
@@ -5544,7 +5611,9 @@ dependencies = [
"tokio-util",
"toml_edit",
"tracing",
"url",
"utils",
"uuid",
]
[[package]]
@@ -5574,6 +5643,7 @@ dependencies = [
"js-sys",
"log",
"mime",
"mime_guess",
"once_cell",
"percent-encoding",
"pin-project-lite",
@@ -7597,6 +7667,16 @@ dependencies = [
"tracing-subscriber",
]
[[package]]
name = "tracing-futures"
version = "0.2.5"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "97d095ae15e245a057c8e8451bab9b3ee1e1f68e9ba2b4fbc18d0ac5237835f2"
dependencies = [
"pin-project",
"tracing",
]
[[package]]
name = "tracing-log"
version = "0.2.0"
@@ -7750,6 +7830,12 @@ dependencies = [
"libc",
]
[[package]]
name = "unicase"
version = "2.8.1"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "75b844d17643ee918803943289730bec8aac480150456169e647ed0b576ba539"
[[package]]
name = "unicode-bidi"
version = "0.3.17"

View File

@@ -40,7 +40,7 @@ members = [
"libs/proxy/postgres-protocol2",
"libs/proxy/postgres-types2",
"libs/proxy/tokio-postgres2",
"endpoint_storage",
"object_storage",
]
[workspace.package]

View File

@@ -89,7 +89,7 @@ RUN set -e \
--bin storage_broker \
--bin storage_controller \
--bin proxy \
--bin endpoint_storage \
--bin object_storage \
--bin neon_local \
--bin storage_scrubber \
--locked --release
@@ -122,7 +122,7 @@ COPY --from=build --chown=neon:neon /home/nonroot/target/release/safekeeper
COPY --from=build --chown=neon:neon /home/nonroot/target/release/storage_broker /usr/local/bin
COPY --from=build --chown=neon:neon /home/nonroot/target/release/storage_controller /usr/local/bin
COPY --from=build --chown=neon:neon /home/nonroot/target/release/proxy /usr/local/bin
COPY --from=build --chown=neon:neon /home/nonroot/target/release/endpoint_storage /usr/local/bin
COPY --from=build --chown=neon:neon /home/nonroot/target/release/object_storage /usr/local/bin
COPY --from=build --chown=neon:neon /home/nonroot/target/release/neon_local /usr/local/bin
COPY --from=build --chown=neon:neon /home/nonroot/target/release/storage_scrubber /usr/local/bin

View File

@@ -1677,7 +1677,7 @@ RUN set -e \
&& apt clean && rm -rf /var/lib/apt/lists/*
# Use `dist_man_MANS=` to skip manpage generation (which requires python3/pandoc)
ENV PGBOUNCER_TAG=pgbouncer_1_24_1
ENV PGBOUNCER_TAG=pgbouncer_1_22_1
RUN set -e \
&& git clone --recurse-submodules --depth 1 --branch ${PGBOUNCER_TAG} https://github.com/pgbouncer/pgbouncer.git pgbouncer \
&& cd pgbouncer \

View File

@@ -11,14 +11,6 @@ index bf6edcb..89b4c7f 100644
USE_PGXS = 1 # use pgxs if not in contrib directory
PGXS := $(shell $(PG_CONFIG) --pgxs)
diff --git a/regress/expected/init-extension.out b/regress/expected/init-extension.out
index 9f2e171..f6e4f8d 100644
--- a/regress/expected/init-extension.out
+++ b/regress/expected/init-extension.out
@@ -1,3 +1,2 @@
SET client_min_messages = warning;
CREATE EXTENSION pg_repack;
-RESET client_min_messages;
diff --git a/regress/expected/nosuper.out b/regress/expected/nosuper.out
index 8d0a94e..63b68bf 100644
--- a/regress/expected/nosuper.out
@@ -50,14 +42,6 @@ index 8d0a94e..63b68bf 100644
INFO: repacking table "public.tbl_cluster"
ERROR: query failed: ERROR: current transaction is aborted, commands ignored until end of transaction block
DETAIL: query was: RESET lock_timeout
diff --git a/regress/sql/init-extension.sql b/regress/sql/init-extension.sql
index 9f2e171..f6e4f8d 100644
--- a/regress/sql/init-extension.sql
+++ b/regress/sql/init-extension.sql
@@ -1,3 +1,2 @@
SET client_min_messages = warning;
CREATE EXTENSION pg_repack;
-RESET client_min_messages;
diff --git a/regress/sql/nosuper.sql b/regress/sql/nosuper.sql
index 072f0fa..dbe60f8 100644
--- a/regress/sql/nosuper.sql

View File

@@ -18,11 +18,12 @@ use anyhow::{Context, Result, anyhow, bail};
use clap::Parser;
use compute_api::spec::ComputeMode;
use control_plane::endpoint::ComputeControlPlane;
use control_plane::endpoint_storage::{ENDPOINT_STORAGE_DEFAULT_PORT, EndpointStorage};
use control_plane::local_env::{
EndpointStorageConf, InitForceMode, LocalEnv, NeonBroker, NeonLocalInitConf,
NeonLocalInitPageserverConf, SafekeeperConf,
InitForceMode, LocalEnv, NeonBroker, NeonLocalInitConf, NeonLocalInitPageserverConf,
ObjectStorageConf, SafekeeperConf,
};
use control_plane::object_storage::OBJECT_STORAGE_DEFAULT_PORT;
use control_plane::object_storage::ObjectStorage;
use control_plane::pageserver::PageServerNode;
use control_plane::safekeeper::SafekeeperNode;
use control_plane::storage_controller::{
@@ -92,7 +93,7 @@ enum NeonLocalCmd {
#[command(subcommand)]
Safekeeper(SafekeeperCmd),
#[command(subcommand)]
EndpointStorage(EndpointStorageCmd),
ObjectStorage(ObjectStorageCmd),
#[command(subcommand)]
Endpoint(EndpointCmd),
#[command(subcommand)]
@@ -459,14 +460,14 @@ enum SafekeeperCmd {
#[derive(clap::Subcommand)]
#[clap(about = "Manage object storage")]
enum EndpointStorageCmd {
Start(EndpointStorageStartCmd),
Stop(EndpointStorageStopCmd),
enum ObjectStorageCmd {
Start(ObjectStorageStartCmd),
Stop(ObjectStorageStopCmd),
}
#[derive(clap::Args)]
#[clap(about = "Start object storage")]
struct EndpointStorageStartCmd {
struct ObjectStorageStartCmd {
#[clap(short = 't', long, help = "timeout until we fail the command")]
#[arg(default_value = "10s")]
start_timeout: humantime::Duration,
@@ -474,7 +475,7 @@ struct EndpointStorageStartCmd {
#[derive(clap::Args)]
#[clap(about = "Stop object storage")]
struct EndpointStorageStopCmd {
struct ObjectStorageStopCmd {
#[arg(value_enum, default_value = "fast")]
#[clap(
short = 'm',
@@ -796,9 +797,7 @@ fn main() -> Result<()> {
}
NeonLocalCmd::StorageBroker(subcmd) => rt.block_on(handle_storage_broker(&subcmd, env)),
NeonLocalCmd::Safekeeper(subcmd) => rt.block_on(handle_safekeeper(&subcmd, env)),
NeonLocalCmd::EndpointStorage(subcmd) => {
rt.block_on(handle_endpoint_storage(&subcmd, env))
}
NeonLocalCmd::ObjectStorage(subcmd) => rt.block_on(handle_object_storage(&subcmd, env)),
NeonLocalCmd::Endpoint(subcmd) => rt.block_on(handle_endpoint(&subcmd, env)),
NeonLocalCmd::Mappings(subcmd) => handle_mappings(&subcmd, env),
};
@@ -1015,8 +1014,8 @@ fn handle_init(args: &InitCmdArgs) -> anyhow::Result<LocalEnv> {
}
})
.collect(),
endpoint_storage: EndpointStorageConf {
port: ENDPOINT_STORAGE_DEFAULT_PORT,
object_storage: ObjectStorageConf {
port: OBJECT_STORAGE_DEFAULT_PORT,
},
pg_distrib_dir: None,
neon_distrib_dir: None,
@@ -1736,15 +1735,12 @@ async fn handle_safekeeper(subcmd: &SafekeeperCmd, env: &local_env::LocalEnv) ->
Ok(())
}
async fn handle_endpoint_storage(
subcmd: &EndpointStorageCmd,
env: &local_env::LocalEnv,
) -> Result<()> {
use EndpointStorageCmd::*;
let storage = EndpointStorage::from_env(env);
async fn handle_object_storage(subcmd: &ObjectStorageCmd, env: &local_env::LocalEnv) -> Result<()> {
use ObjectStorageCmd::*;
let storage = ObjectStorage::from_env(env);
// In tests like test_forward_compatibility or test_graceful_cluster_restart
// old neon binaries (without endpoint_storage) are present
// old neon binaries (without object_storage) are present
if !storage.bin.exists() {
eprintln!(
"{} binary not found. Ignore if this is a compatibility test",
@@ -1754,13 +1750,13 @@ async fn handle_endpoint_storage(
}
match subcmd {
Start(EndpointStorageStartCmd { start_timeout }) => {
Start(ObjectStorageStartCmd { start_timeout }) => {
if let Err(e) = storage.start(start_timeout).await {
eprintln!("endpoint_storage start failed: {e}");
eprintln!("object_storage start failed: {e}");
exit(1);
}
}
Stop(EndpointStorageStopCmd { stop_mode }) => {
Stop(ObjectStorageStopCmd { stop_mode }) => {
let immediate = match stop_mode {
StopMode::Fast => false,
StopMode::Immediate => true,
@@ -1870,10 +1866,10 @@ async fn handle_start_all_impl(
}
js.spawn(async move {
EndpointStorage::from_env(env)
ObjectStorage::from_env(env)
.start(&retry_timeout)
.await
.map_err(|e| e.context("start endpoint_storage"))
.map_err(|e| e.context("start object_storage"))
});
})();
@@ -1972,9 +1968,9 @@ async fn try_stop_all(env: &local_env::LocalEnv, immediate: bool) {
}
}
let storage = EndpointStorage::from_env(env);
let storage = ObjectStorage::from_env(env);
if let Err(e) = storage.stop(immediate) {
eprintln!("endpoint_storage stop failed: {:#}", e);
eprintln!("object_storage stop failed: {:#}", e);
}
for ps_conf in &env.pageservers {

View File

@@ -9,8 +9,8 @@
mod background_process;
pub mod broker;
pub mod endpoint;
pub mod endpoint_storage;
pub mod local_env;
pub mod object_storage;
pub mod pageserver;
pub mod postgresql_conf;
pub mod safekeeper;

View File

@@ -19,7 +19,7 @@ use serde::{Deserialize, Serialize};
use utils::auth::encode_from_key_file;
use utils::id::{NodeId, TenantId, TenantTimelineId, TimelineId};
use crate::endpoint_storage::{ENDPOINT_STORAGE_REMOTE_STORAGE_DIR, EndpointStorage};
use crate::object_storage::{OBJECT_STORAGE_REMOTE_STORAGE_DIR, ObjectStorage};
use crate::pageserver::{PAGESERVER_REMOTE_STORAGE_DIR, PageServerNode};
use crate::safekeeper::SafekeeperNode;
@@ -72,7 +72,7 @@ pub struct LocalEnv {
pub safekeepers: Vec<SafekeeperConf>,
pub endpoint_storage: EndpointStorageConf,
pub object_storage: ObjectStorageConf,
// Control plane upcall API for pageserver: if None, we will not run storage_controller If set, this will
// be propagated into each pageserver's configuration.
@@ -110,7 +110,7 @@ pub struct OnDiskConfig {
)]
pub pageservers: Vec<PageServerConf>,
pub safekeepers: Vec<SafekeeperConf>,
pub endpoint_storage: EndpointStorageConf,
pub object_storage: ObjectStorageConf,
pub control_plane_api: Option<Url>,
pub control_plane_hooks_api: Option<Url>,
pub control_plane_compute_hook_api: Option<Url>,
@@ -144,7 +144,7 @@ pub struct NeonLocalInitConf {
pub storage_controller: Option<NeonStorageControllerConf>,
pub pageservers: Vec<NeonLocalInitPageserverConf>,
pub safekeepers: Vec<SafekeeperConf>,
pub endpoint_storage: EndpointStorageConf,
pub object_storage: ObjectStorageConf,
pub control_plane_api: Option<Url>,
pub control_plane_hooks_api: Option<Url>,
pub generate_local_ssl_certs: bool,
@@ -152,7 +152,7 @@ pub struct NeonLocalInitConf {
#[derive(Serialize, Default, Deserialize, PartialEq, Eq, Clone, Debug)]
#[serde(default)]
pub struct EndpointStorageConf {
pub struct ObjectStorageConf {
pub port: u16,
}
@@ -413,8 +413,8 @@ impl LocalEnv {
self.pg_dir(pg_version, "lib")
}
pub fn endpoint_storage_bin(&self) -> PathBuf {
self.neon_distrib_dir.join("endpoint_storage")
pub fn object_storage_bin(&self) -> PathBuf {
self.neon_distrib_dir.join("object_storage")
}
pub fn pageserver_bin(&self) -> PathBuf {
@@ -450,8 +450,8 @@ impl LocalEnv {
self.base_data_dir.join("safekeepers").join(data_dir_name)
}
pub fn endpoint_storage_data_dir(&self) -> PathBuf {
self.base_data_dir.join("endpoint_storage")
pub fn object_storage_data_dir(&self) -> PathBuf {
self.base_data_dir.join("object_storage")
}
pub fn get_pageserver_conf(&self, id: NodeId) -> anyhow::Result<&PageServerConf> {
@@ -615,7 +615,7 @@ impl LocalEnv {
control_plane_compute_hook_api: _,
branch_name_mappings,
generate_local_ssl_certs,
endpoint_storage,
object_storage,
} = on_disk_config;
LocalEnv {
base_data_dir: repopath.to_owned(),
@@ -632,7 +632,7 @@ impl LocalEnv {
control_plane_hooks_api,
branch_name_mappings,
generate_local_ssl_certs,
endpoint_storage,
object_storage,
}
};
@@ -742,7 +742,7 @@ impl LocalEnv {
control_plane_compute_hook_api: None,
branch_name_mappings: self.branch_name_mappings.clone(),
generate_local_ssl_certs: self.generate_local_ssl_certs,
endpoint_storage: self.endpoint_storage.clone(),
object_storage: self.object_storage.clone(),
},
)
}
@@ -849,7 +849,7 @@ impl LocalEnv {
control_plane_api,
generate_local_ssl_certs,
control_plane_hooks_api,
endpoint_storage,
object_storage,
} = conf;
// Find postgres binaries.
@@ -901,7 +901,7 @@ impl LocalEnv {
control_plane_hooks_api,
branch_name_mappings: Default::default(),
generate_local_ssl_certs,
endpoint_storage,
object_storage,
};
if generate_local_ssl_certs {
@@ -929,13 +929,13 @@ impl LocalEnv {
.context("pageserver init failed")?;
}
EndpointStorage::from_env(&env)
ObjectStorage::from_env(&env)
.init()
.context("object storage init failed")?;
// setup remote remote location for default LocalFs remote storage
std::fs::create_dir_all(env.base_data_dir.join(PAGESERVER_REMOTE_STORAGE_DIR))?;
std::fs::create_dir_all(env.base_data_dir.join(ENDPOINT_STORAGE_REMOTE_STORAGE_DIR))?;
std::fs::create_dir_all(env.base_data_dir.join(OBJECT_STORAGE_REMOTE_STORAGE_DIR))?;
env.persist_config()
}

View File

@@ -1,33 +1,34 @@
use crate::background_process::{self, start_process, stop_process};
use crate::local_env::LocalEnv;
use anyhow::anyhow;
use anyhow::{Context, Result};
use camino::Utf8PathBuf;
use std::io::Write;
use std::time::Duration;
/// Directory within .neon which will be used by default for LocalFs remote storage.
pub const ENDPOINT_STORAGE_REMOTE_STORAGE_DIR: &str = "local_fs_remote_storage/endpoint_storage";
pub const ENDPOINT_STORAGE_DEFAULT_PORT: u16 = 9993;
pub const OBJECT_STORAGE_REMOTE_STORAGE_DIR: &str = "local_fs_remote_storage/object_storage";
pub const OBJECT_STORAGE_DEFAULT_PORT: u16 = 9993;
pub struct EndpointStorage {
pub struct ObjectStorage {
pub bin: Utf8PathBuf,
pub data_dir: Utf8PathBuf,
pub pemfile: Utf8PathBuf,
pub port: u16,
}
impl EndpointStorage {
pub fn from_env(env: &LocalEnv) -> EndpointStorage {
EndpointStorage {
bin: Utf8PathBuf::from_path_buf(env.endpoint_storage_bin()).unwrap(),
data_dir: Utf8PathBuf::from_path_buf(env.endpoint_storage_data_dir()).unwrap(),
impl ObjectStorage {
pub fn from_env(env: &LocalEnv) -> ObjectStorage {
ObjectStorage {
bin: Utf8PathBuf::from_path_buf(env.object_storage_bin()).unwrap(),
data_dir: Utf8PathBuf::from_path_buf(env.object_storage_data_dir()).unwrap(),
pemfile: Utf8PathBuf::from_path_buf(env.public_key_path.clone()).unwrap(),
port: env.endpoint_storage.port,
port: env.object_storage.port,
}
}
fn config_path(&self) -> Utf8PathBuf {
self.data_dir.join("endpoint_storage.json")
self.data_dir.join("object_storage.json")
}
fn listen_addr(&self) -> Utf8PathBuf {
@@ -48,7 +49,7 @@ impl EndpointStorage {
let cfg = Cfg {
listen: self.listen_addr(),
pemfile: parent.join(self.pemfile.clone()),
local_path: parent.join(ENDPOINT_STORAGE_REMOTE_STORAGE_DIR),
local_path: parent.join(OBJECT_STORAGE_REMOTE_STORAGE_DIR),
r#type: "LocalFs".to_string(),
};
std::fs::create_dir_all(self.config_path().parent().unwrap())?;
@@ -58,19 +59,24 @@ impl EndpointStorage {
}
pub async fn start(&self, retry_timeout: &Duration) -> Result<()> {
println!("Starting endpoint_storage at {}", self.listen_addr());
println!("Starting s3 proxy at {}", self.listen_addr());
std::io::stdout().flush().context("flush stdout")?;
let process_status_check = || async {
let res = reqwest::Client::new().get(format!("http://{}/metrics", self.listen_addr()));
match res.send().await {
Ok(res) => Ok(res.status().is_success()),
Err(_) => Ok(false),
tokio::time::sleep(Duration::from_millis(500)).await;
let res = reqwest::Client::new()
.get(format!("http://{}/metrics", self.listen_addr()))
.send()
.await;
match res {
Ok(response) if response.status().is_success() => Ok(true),
Ok(_) => Err(anyhow!("Failed to query /metrics")),
Err(e) => Err(anyhow!("Failed to check node status: {e}")),
}
};
let res = start_process(
"endpoint_storage",
"object_storage",
&self.data_dir.clone().into_std_path_buf(),
&self.bin.clone().into_std_path_buf(),
vec![self.config_path().to_string()],
@@ -88,14 +94,14 @@ impl EndpointStorage {
}
pub fn stop(&self, immediate: bool) -> anyhow::Result<()> {
stop_process(immediate, "endpoint_storage", &self.pid_file())
stop_process(immediate, "object_storage", &self.pid_file())
}
fn log_file(&self) -> Utf8PathBuf {
self.data_dir.join("endpoint_storage.log")
self.data_dir.join("object_storage.log")
}
fn pid_file(&self) -> Utf8PathBuf {
self.data_dir.join("endpoint_storage.pid")
self.data_dir.join("object_storage.pid")
}
}

View File

@@ -1,53 +0,0 @@
# Architecture Naming Scheme
## Summary
Neon computes are going to support multiple CPU architectures, including 64-bit
x86 and ARM. Architecture naming schemes are fairly inconsistent, at least when
it comes to these two architectures in particular. Sometimes 64-bit x86 is known
as `amd64` and at other times `x86_64`. For 64-bit ARM, it's a similar story
with `arm64` and `aarch64`.
## Motivation
Consistency when referring to these architectures across the platform in
important not only so that everyone is on the same page, but also because we
have architecture-dependent portions of the platform. One notable example is
remote extensions. Remote extensions need to be compiled and packaged into
compressed tarballs that are then downloaded by computes on demand. Downloading
a tarball for the wrong architecture will cause `dlopen(3)` to fail when
Postgres attempts to load the library. The compressed tarballs are located in an
S3 bucket with object keys of the form
`$BUILD_TAG/$ARCH/$PG_VERSION_NUM/${EXTENSION_NAME}.tar.zst` to mitigate the
potential for failure. The build and deployment pipeline for remote extensions
needs to be in lock step with the code in the compute that actually fetches
remote extensions.
## Impacted Components
- Control Plane when persisting the target architecture in compute flags if
specified.
- Austocaling when scheduling compute pods.
- `compute_ctl` when downloading remote extensions.
- Build and deployment pipeline for remote extensions.
## Prior Art for CPU Architecture Names
- [Rust](https://doc.rust-lang.org/std/env/consts/constant.ARCH.html)
- `x86_64`
- `aarch64`
- [Go](https://pkg.go.dev/internal/goarch#pkg-constants)
- `amd64`
- `arm64`
- Kubernetes
- Because Kubernetes is written in Go, it inherits the naming scheme.
Going all in on either the Rust naming scheme or the Go naming scheme can save
us a branch when getting the architecture, so we should pick one and force the
other side to conform.
## Decision
The heart of our platform is Kubernetes. We will inherit the Go naming
scheme just like Kubernetes. This makes things easy for the autoscaling team
when they do pod scheduling on nodes.

View File

@@ -242,22 +242,13 @@ impl RemoteExtSpec {
match self.extension_data.get(real_ext_name) {
Some(_ext_data) => {
// We have decided to use the Go naming convention due to Kubernetes.
let arch = match std::env::consts::ARCH {
"x86_64" => "amd64",
"aarch64" => "arm64",
arch => arch,
};
// Construct the path to the extension archive
// BUILD_TAG/PG_MAJOR_VERSION/extensions/EXTENSION_NAME.tar.zst
//
// Keep it in sync with path generation in
// https://github.com/neondatabase/build-custom-extensions/tree/main
let archive_path_str = format!(
"{build_tag}/{arch}/{pg_major_version}/extensions/{real_ext_name}.tar.zst"
);
let archive_path_str =
format!("{build_tag}/{pg_major_version}/extensions/{real_ext_name}.tar.zst");
Ok((
real_ext_name.to_string(),
RemotePath::from_string(&archive_path_str)?,

View File

@@ -18,7 +18,8 @@ camino = { workspace = true, features = ["serde1"] }
humantime-serde.workspace = true
hyper = { workspace = true, features = ["client"] }
futures.workspace = true
reqwest.workspace = true
reqwest = { workspace = true, features = ["multipart", "stream"] }
chrono = { version = "0.4", default-features = false, features = ["clock"] }
serde.workspace = true
serde_json.workspace = true
tokio = { workspace = true, features = ["sync", "fs", "io-util"] }
@@ -40,6 +41,10 @@ http-types.workspace = true
http-body-util.workspace = true
itertools.workspace = true
sync_wrapper = { workspace = true, features = ["futures"] }
gcp_auth = "0.12.3"
url.workspace = true
http.workspace = true
uuid.workspace = true
[dev-dependencies]
camino-tempfile.workspace = true

View File

@@ -41,6 +41,7 @@ impl RemoteStorageKind {
RemoteStorageKind::LocalFs { .. } => None,
RemoteStorageKind::AwsS3(config) => Some(&config.bucket_name),
RemoteStorageKind::AzureContainer(config) => Some(&config.container_name),
RemoteStorageKind::GCS(config) => Some(&config.bucket_name),
}
}
}
@@ -51,6 +52,7 @@ impl RemoteStorageConfig {
match &self.storage {
RemoteStorageKind::LocalFs { .. } => DEFAULT_REMOTE_STORAGE_LOCALFS_CONCURRENCY_LIMIT,
RemoteStorageKind::AwsS3(c) => c.concurrency_limit.into(),
RemoteStorageKind::GCS(c) => c.concurrency_limit.into(),
RemoteStorageKind::AzureContainer(c) => c.concurrency_limit.into(),
}
}
@@ -85,6 +87,9 @@ pub enum RemoteStorageKind {
/// Azure Blob based storage, storing all files in the container
/// specified by the config
AzureContainer(AzureConfig),
/// Google Cloud based storage, storing all files in the GCS bucket
/// specified by the config
GCS(GCSConfig),
}
/// AWS S3 bucket coordinates and access credentials to manage the bucket contents (read and write).
@@ -154,6 +159,32 @@ impl Debug for S3Config {
}
}
#[derive(Clone, PartialEq, Eq, Deserialize, Serialize)]
pub struct GCSConfig {
/// Name of the bucket to connect to.
pub bucket_name: String,
/// A "subfolder" in the bucket, to use the same bucket separately by multiple remote storage users at once.
pub prefix_in_bucket: Option<String>,
#[serde(default = "default_remote_storage_s3_concurrency_limit")]
pub concurrency_limit: NonZeroUsize,
#[serde(default = "default_max_keys_per_list_response")]
pub max_keys_per_list_response: Option<i32>,
}
impl Debug for GCSConfig {
fn fmt(&self, f: &mut std::fmt::Formatter<'_>) -> std::fmt::Result {
f.debug_struct("GCSConfig")
.field("bucket_name", &self.bucket_name)
.field("prefix_in_bucket", &self.prefix_in_bucket)
.field("concurrency_limit", &self.concurrency_limit)
.field(
"max_keys_per_list_response",
&self.max_keys_per_list_response,
)
.finish()
}
}
/// Azure bucket coordinates and access credentials to manage the bucket contents (read and write).
#[derive(Clone, PartialEq, Eq, Serialize, Deserialize)]
pub struct AzureConfig {
@@ -268,6 +299,30 @@ timeout = '5s'";
);
}
#[test]
fn test_gcs_parsing() {
let toml = "\
bucket_name = 'foo-bar'
prefix_in_bucket = '/pageserver'
";
let config = parse(toml).unwrap();
assert_eq!(
config,
RemoteStorageConfig {
storage: RemoteStorageKind::GCS(GCSConfig {
bucket_name: "foo-bar".into(),
prefix_in_bucket: Some("pageserver/".into()),
max_keys_per_list_response: DEFAULT_MAX_KEYS_PER_LIST_RESPONSE,
concurrency_limit: std::num::NonZero::new(100).unwrap(),
}),
timeout: Duration::from_secs(120),
small_timeout: RemoteStorageConfig::DEFAULT_SMALL_TIMEOUT
}
);
}
#[test]
fn test_s3_parsing() {
let toml = "\

View File

@@ -0,0 +1,978 @@
#![allow(dead_code)]
#![allow(unused)]
use crate::config::GCSConfig;
use crate::error::Cancelled;
pub(super) use crate::metrics::RequestKind;
use crate::metrics::{AttemptOutcome, start_counting_cancelled_wait, start_measuring_requests};
use crate::{
ConcurrencyLimiter, Download, DownloadError, DownloadOpts, GCS_SCOPES, Listing, ListingMode,
ListingObject, MAX_KEYS_PER_DELETE_GCS, REMOTE_STORAGE_PREFIX_SEPARATOR, RemotePath,
RemoteStorage, StorageMetadata, TimeTravelError, TimeoutOrCancel,
};
use anyhow::Context;
use azure_core::Etag;
use bytes::Bytes;
use bytes::BytesMut;
use chrono::DateTime;
use futures::stream::Stream;
use futures::stream::TryStreamExt;
use futures_util::StreamExt;
use gcp_auth::{Token, TokenProvider};
use http::Method;
use http::StatusCode;
use reqwest::{Client, header};
use scopeguard::ScopeGuard;
use serde::{Deserialize, Serialize};
use std::collections::HashMap;
use std::fmt::Debug;
use std::num::NonZeroU32;
use std::pin::{Pin, pin};
use std::sync::Arc;
use std::time::Duration;
use std::time::SystemTime;
use tokio_util::codec::{BytesCodec, FramedRead};
use tokio_util::sync::CancellationToken;
use tracing;
use url::Url;
use uuid::Uuid;
// ---------
pub struct GCSBucket {
token_provider: Arc<dyn TokenProvider>,
bucket_name: String,
prefix_in_bucket: Option<String>,
max_keys_per_list_response: Option<i32>,
concurrency_limiter: ConcurrencyLimiter,
pub timeout: Duration,
}
struct GetObjectRequest {
bucket: String,
key: String,
etag: Option<String>,
range: Option<String>,
}
// ---------
impl GCSBucket {
pub async fn new(remote_storage_config: &GCSConfig, timeout: Duration) -> anyhow::Result<Self> {
tracing::debug!(
"creating remote storage for gcs bucket {}",
remote_storage_config.bucket_name
);
// clean up 'prefix_in_bucket' if user provides '/pageserver' or 'pageserver/'
let prefix_in_bucket = remote_storage_config
.prefix_in_bucket
.as_deref()
.map(|prefix| {
let mut prefix = prefix;
while prefix.starts_with(REMOTE_STORAGE_PREFIX_SEPARATOR) {
prefix = &prefix[1..];
}
let mut prefix = prefix.to_string();
if prefix.ends_with(REMOTE_STORAGE_PREFIX_SEPARATOR) {
prefix.pop();
}
prefix
});
// get GOOGLE_APPLICATION_CREDENTIALS
let provider = gcp_auth::provider().await?;
Ok(GCSBucket {
token_provider: Arc::clone(&provider),
bucket_name: remote_storage_config.bucket_name.clone(),
prefix_in_bucket,
timeout,
max_keys_per_list_response: remote_storage_config.max_keys_per_list_response,
concurrency_limiter: ConcurrencyLimiter::new(
remote_storage_config.concurrency_limit.get(),
),
})
}
// convert `RemotePath` -> `String`
pub fn relative_path_to_gcs_object(&self, path: &RemotePath) -> String {
let path_string = path.get_path().as_str();
match &self.prefix_in_bucket {
Some(prefix) => prefix.clone() + "/" + path_string,
None => path_string.to_string(),
}
}
// convert `String` -> `RemotePath`
pub fn gcs_object_to_relative_path(&self, key: &str) -> RemotePath {
let relative_path =
match key.strip_prefix(self.prefix_in_bucket.as_deref().unwrap_or_default()) {
Some(stripped) => stripped,
// we rely on GCS to return properly prefixed paths
// for requests with a certain prefix
None => panic!(
"Key {} does not start with bucket prefix {:?}",
key, self.prefix_in_bucket
),
};
RemotePath(
relative_path
.split(REMOTE_STORAGE_PREFIX_SEPARATOR)
.collect(),
)
}
pub fn bucket_name(&self) -> &str {
&self.bucket_name
}
fn max_keys_per_delete(&self) -> usize {
MAX_KEYS_PER_DELETE_GCS
}
async fn permit(
&self,
kind: RequestKind,
cancel: &CancellationToken,
) -> Result<tokio::sync::SemaphorePermit<'_>, Cancelled> {
let started_at = start_counting_cancelled_wait(kind);
let acquire = self.concurrency_limiter.acquire(kind);
let permit = tokio::select! {
permit = acquire => permit.expect("semaphore is never closed"),
_ = cancel.cancelled() => return Err(Cancelled),
};
let started_at = ScopeGuard::into_inner(started_at);
crate::metrics::BUCKET_METRICS
.wait_seconds
.observe_elapsed(kind, started_at);
Ok(permit)
}
async fn owned_permit(
&self,
kind: RequestKind,
cancel: &CancellationToken,
) -> Result<tokio::sync::OwnedSemaphorePermit, Cancelled> {
let started_at = start_counting_cancelled_wait(kind);
let acquire = self.concurrency_limiter.acquire_owned(kind);
let permit = tokio::select! {
permit = acquire => permit.expect("semaphore is never closed"),
_ = cancel.cancelled() => return Err(Cancelled),
};
let started_at = ScopeGuard::into_inner(started_at);
crate::metrics::BUCKET_METRICS
.wait_seconds
.observe_elapsed(kind, started_at);
Ok(permit)
}
async fn put_object(
&self,
byte_stream: impl Stream<Item = std::io::Result<Bytes>> + Send + Sync + 'static,
fs_size: usize,
to: &RemotePath,
cancel: &CancellationToken,
) -> anyhow::Result<()> {
// https://cloud.google.com/storage/docs/xml-api/reference-headers#chunked
let mut headers = header::HeaderMap::new();
headers.insert(
header::TRANSFER_ENCODING,
header::HeaderValue::from_static("chunked"),
);
// TODO Check if we need type 'multipart/related' file to attach metadata like Neon's S3
// `.upload()` does.
// https://cloud.google.com/storage/docs/uploading-objects#uploading-an-object
let upload_uri = format!(
"https://storage.googleapis.com/upload/storage/v1/b/{}/o/?uploadType=media&name={}",
self.bucket_name.clone(),
self.relative_path_to_gcs_object(to).trim_start_matches("/")
);
let upload = Client::new()
.post(upload_uri)
.body(reqwest::Body::wrap_stream(byte_stream))
.headers(headers)
.bearer_auth(self.token_provider.token(GCS_SCOPES).await?.as_str())
.send();
// We await it in a race against the Tokio timeout
let upload = tokio::time::timeout(self.timeout, upload);
let res = tokio::select! {
res = upload => res,
_ = cancel.cancelled() => return Err(TimeoutOrCancel::Cancel.into()),
};
match res {
Ok(Ok(res)) => {
if !res.status().is_success() {
match res.status() {
StatusCode::NOT_FOUND => {
return Err(anyhow::anyhow!("GCS error: not found \n\t {:?}", res));
}
_ => {
return Err(anyhow::anyhow!(
"GCS PUT response contained no response body \n\t {:?}",
res
));
}
}
} else {
Ok(())
}
}
Ok(Err(reqw)) => Err(reqw.into()),
Err(_timeout) => Err(TimeoutOrCancel::Timeout.into()),
}
}
async fn copy(
&self,
from: String,
to: String,
cancel: &CancellationToken,
) -> anyhow::Result<()> {
let kind = RequestKind::Copy;
let _permit = self.permit(kind, cancel).await?;
let timeout = tokio::time::sleep(self.timeout);
let started_at = start_measuring_requests(kind);
let copy_uri = format!(
"https://storage.googleapis.com/storage/v1/b/{}/o/{}/copyTo/b/{}/o/{}",
self.bucket_name.clone(),
&from,
self.bucket_name.clone(),
&to
);
let op = Client::new()
.post(copy_uri)
.bearer_auth(self.token_provider.token(GCS_SCOPES).await?.as_str())
.send();
let res = tokio::select! {
res = op => res,
_ = timeout => return Err(TimeoutOrCancel::Timeout.into()),
_ = cancel.cancelled() => return Err(TimeoutOrCancel::Cancel.into()),
};
let started_at = ScopeGuard::into_inner(started_at);
crate::metrics::BUCKET_METRICS
.req_seconds
.observe_elapsed(kind, &res, started_at);
res?;
Ok(())
}
async fn delete_oids(
&self,
delete_objects: &[String],
cancel: &CancellationToken,
_permit: &tokio::sync::SemaphorePermit<'_>,
) -> anyhow::Result<()> {
let kind = RequestKind::Delete;
let mut cancel = std::pin::pin!(cancel.cancelled());
for chunk in delete_objects.chunks(MAX_KEYS_PER_DELETE_GCS) {
let started_at = start_measuring_requests(kind);
// Use this to report keys that didn't delete based on 'content_id'
let mut delete_objects_status = HashMap::new();
let mut form = reqwest::multipart::Form::new();
let bulk_uri = "https://storage.googleapis.com/batch/storage/v1";
for (index, path) in delete_objects.iter().enumerate() {
delete_objects_status.insert(index + 1, path.clone());
let path_to_delete: String =
url::form_urlencoded::byte_serialize(path.trim_start_matches("/").as_bytes())
.collect();
let delete_req = format!(
"
DELETE /storage/v1/b/{}/o/{} HTTP/1.1\r\n\
Content-Type: application/json\r\n\
accept: application/json\r\n\
content-length: 0\r\n
",
self.bucket_name.clone(),
path_to_delete
)
.trim()
.to_string();
let content_id = format!("<{}+{}>", Uuid::new_v4(), index + 1);
let mut part_headers = header::HeaderMap::new();
part_headers.insert(
header::CONTENT_TYPE,
header::HeaderValue::from_static("application/http"),
);
part_headers.insert(
header::TRANSFER_ENCODING,
header::HeaderValue::from_static("binary"),
);
part_headers.insert(
header::HeaderName::from_static("content-id"),
header::HeaderValue::from_str(&content_id)?,
);
let part = reqwest::multipart::Part::text(delete_req).headers(part_headers);
form = form.part(format!("request-{}", index), part);
}
let mut headers = header::HeaderMap::new();
headers.insert(
header::CONTENT_TYPE,
header::HeaderValue::from_str(&format!(
"multipart/mixed; boundary={}",
form.boundary()
))?,
);
let req = Client::new()
.post(bulk_uri)
.bearer_auth(self.token_provider.token(GCS_SCOPES).await?.as_str())
.multipart(form)
.headers(headers)
.send();
let resp = tokio::select! {
resp = req => resp,
_ = tokio::time::sleep(self.timeout) => return Err(TimeoutOrCancel::Timeout.into()),
_ = &mut cancel => return Err(TimeoutOrCancel::Cancel.into()),
};
let started_at = ScopeGuard::into_inner(started_at);
crate::metrics::BUCKET_METRICS
.req_seconds
.observe_elapsed(kind, &resp, started_at);
let resp = resp.context("request deletion")?;
crate::metrics::BUCKET_METRICS
.deleted_objects_total
.inc_by(chunk.len() as u64);
let res_headers = resp.headers().to_owned();
let boundary = res_headers
.get(header::CONTENT_TYPE)
.unwrap()
.to_str()?
.split("=")
.last()
.unwrap();
let res_body = resp.text().await?;
let parsed: HashMap<String, String> = res_body
.split(&format!("--{}", boundary))
.filter_map(|c| {
let mut lines = c.lines();
let id = lines.find_map(|line| {
line.strip_prefix("Content-ID:")
.and_then(|suf| suf.split('+').last())
.and_then(|suf| suf.split('>').next())
.map(|x| x.trim().to_string())
});
let status_code = lines.find_map(|line| {
// Not sure if this protocol version shouldn't be so specific
line.strip_prefix("HTTP/1.1")
.and_then(|x| x.split_whitespace().next())
.map(|x| x.trim().to_string())
});
id.zip(status_code)
})
.collect();
// Gather failures
let errors: HashMap<usize, &String> = parsed
.iter()
.filter_map(|(x, y)| {
if y.chars().next() != Some('2') {
x.parse::<usize>().ok().map(|v| (v, y))
} else {
None
}
})
.collect();
if !errors.is_empty() {
// Report 10 of them like S3
const LOG_UP_TO_N_ERRORS: usize = 10;
for (id, code) in errors.iter().take(LOG_UP_TO_N_ERRORS) {
tracing::warn!(
"DeleteObjects key {} failed with code: {}",
delete_objects_status.get(id).unwrap(),
code
);
}
return Err(anyhow::anyhow!(
"Failed to delete {}/{} objects",
errors.len(),
chunk.len(),
));
}
}
Ok(())
}
async fn list_objects_v2(&self, list_uri: String) -> anyhow::Result<reqwest::RequestBuilder> {
let res = Client::new()
.get(list_uri)
.bearer_auth(self.token_provider.token(GCS_SCOPES).await?.as_str());
Ok(res)
}
// need a 'bucket', a 'key', and a bytes 'range'.
async fn get_object(
&self,
request: GetObjectRequest,
cancel: &CancellationToken,
) -> anyhow::Result<Download, DownloadError> {
let kind = RequestKind::Get;
let permit = self.owned_permit(kind, cancel).await?;
let started_at = start_measuring_requests(kind);
let encoded_path: String =
url::form_urlencoded::byte_serialize(request.key.as_bytes()).collect();
/// We do this in two parts:
/// 1. Serialize the metadata of the first request to get Etag, last modified, etc
/// 2. We do not .await the second request pass on the pinned stream to the 'get_object'
/// caller
// 1. Serialize Metadata in initial request
let metadata_uri_mod = "alt=json";
let download_uri = format!(
"https://storage.googleapis.com/storage/v1/b/{}/o/{}?{}",
self.bucket_name.clone(),
encoded_path,
metadata_uri_mod
);
let res = Client::new()
.get(download_uri)
.bearer_auth(
self.token_provider
.token(GCS_SCOPES)
.await
.map_err(|e: gcp_auth::Error| DownloadError::Other(e.into()))?
.as_str(),
)
.send()
.await
.map_err(|e: reqwest::Error| DownloadError::Other(e.into()))?;
if !res.status().is_success() {
match res.status() {
StatusCode::NOT_FOUND => return Err(DownloadError::NotFound),
_ => {
return Err(DownloadError::Other(anyhow::anyhow!(
"GCS GET resposne contained no response body"
)));
}
}
};
let body = res
.text()
.await
.map_err(|e: reqwest::Error| DownloadError::Other(e.into()))?;
let resp: GCSObject = serde_json::from_str(&body)
.map_err(|e: serde_json::Error| DownloadError::Other(e.into()))?;
// 2. Byte Stream request
let mut headers = header::HeaderMap::new();
headers.insert(header::RANGE, header::HeaderValue::from_static("bytes=0-"));
let encoded_path: String =
url::form_urlencoded::byte_serialize(request.key.as_bytes()).collect();
let stream_uri_mod = "alt=media";
let stream_uri = format!(
"https://storage.googleapis.com/storage/v1/b/{}/o/{}?{}",
self.bucket_name.clone(),
encoded_path,
stream_uri_mod
);
let mut req = Client::new()
.get(stream_uri)
.headers(headers)
.bearer_auth(
self.token_provider
.token(GCS_SCOPES)
.await
.map_err(|e: gcp_auth::Error| DownloadError::Other(e.into()))?
.as_str(),
)
.send();
let get_object = tokio::select! {
res = req => res,
_ = tokio::time::sleep(self.timeout) => return Err(DownloadError::Timeout),
_ = cancel.cancelled() => return Err(DownloadError::Cancelled),
};
let started_at = ScopeGuard::into_inner(started_at);
let object_output = match get_object {
Ok(object_output) => {
if !object_output.status().is_success() {
match object_output.status() {
StatusCode::NOT_FOUND => return Err(DownloadError::NotFound),
_ => {
return Err(DownloadError::Other(anyhow::anyhow!(
"GCS GET resposne contained no response body"
)));
}
}
} else {
object_output
}
}
Err(e) => {
crate::metrics::BUCKET_METRICS.req_seconds.observe_elapsed(
kind,
AttemptOutcome::Err,
started_at,
);
return Err(DownloadError::Other(
anyhow::Error::new(e).context("download s3 object"),
));
}
};
let remaining = self.timeout.saturating_sub(started_at.elapsed());
let metadata = resp.metadata.map(StorageMetadata);
let etag = resp
.etag
.ok_or(DownloadError::Other(anyhow::anyhow!("Missing ETag header")))?
.into();
let last_modified: SystemTime = resp
.updated
.and_then(|s| DateTime::parse_from_rfc3339(&s).ok())
.map(|s| s.into())
.unwrap_or(SystemTime::now());
// But let data stream pass through
Ok(Download {
download_stream: Box::pin(object_output.bytes_stream().map(|item| {
item.map_err(|e: reqwest::Error| std::io::Error::new(std::io::ErrorKind::Other, e))
})),
etag,
last_modified,
metadata,
})
}
}
impl RemoteStorage for GCSBucket {
// ---------------------------------------
// Neon wrappers for GCS client functions
// ---------------------------------------
fn list_streaming(
&self,
prefix: Option<&RemotePath>,
mode: ListingMode,
max_keys: Option<NonZeroU32>,
cancel: &CancellationToken,
) -> impl Stream<Item = Result<Listing, DownloadError>> {
let kind = RequestKind::List;
let mut max_keys = max_keys.map(|mk| mk.get() as i32);
let list_prefix = prefix
.map(|p| self.relative_path_to_gcs_object(p))
.or_else(|| {
self.prefix_in_bucket.clone().map(|mut s| {
s.push(REMOTE_STORAGE_PREFIX_SEPARATOR);
s
})
})
.unwrap();
let request_max_keys = self
.max_keys_per_list_response
.into_iter()
.chain(max_keys.into_iter())
.min()
// https://cloud.google.com/storage/docs/json_api/v1/objects/list?hl=en#parameters
// TODO set this to default
.unwrap_or(1000);
// We pass URI in to `list_objects_v2` as we'll modify it with `NextPageToken`, hence
// `mut`
let mut list_uri = format!(
"https://storage.googleapis.com/storage/v1/b/{}/o?prefix={}&maxResults={}",
self.bucket_name.clone(),
list_prefix,
request_max_keys,
);
// on ListingMode:
// https://github.com/neondatabase/neon/blob/edc11253b65e12a10843711bd88ad277511396d7/libs/remote_storage/src/lib.rs#L158C1-L164C2
if let ListingMode::WithDelimiter = mode {
list_uri.push_str(&format!(
"&delimiter={}",
REMOTE_STORAGE_PREFIX_SEPARATOR.to_string()
));
}
async_stream::stream! {
let mut continuation_token = None;
'outer: loop {
let started_at = start_measuring_requests(kind);
let request = self.list_objects_v2(list_uri.clone())
.await
.map_err(DownloadError::Other)?
.send();
// this is like `await`
let response = tokio::select! {
res = request => Ok(res),
_ = tokio::time::sleep(self.timeout) => Err(DownloadError::Timeout),
_ = cancel.cancelled() => Err(DownloadError::Cancelled),
}?;
// just mapping our `Result' error variant's type.
let response = response
.context("Failed to list GCS prefixes")
.map_err(DownloadError::Other);
let started_at = ScopeGuard::into_inner(started_at);
crate::metrics::BUCKET_METRICS
.req_seconds
.observe_elapsed(kind, &response, started_at);
let response = match response {
Ok(response) => response,
Err(e) => {
// The error is potentially retryable, so we must rewind the loop after yielding.
yield Err(e);
continue 'outer;
},
};
let body = response.text()
.await
.map_err(|e: reqwest::Error| DownloadError::Other(e.into()))?;
let resp: GCSListResponse = serde_json::from_str(&body).map_err(|e: serde_json::Error| DownloadError::Other(e.into()))?;
let prefixes = resp.common_prefixes();
let keys = resp.contents();
tracing::debug!("list: {} prefixes, {} keys", prefixes.len(), keys.len());
let mut result = Listing::default();
for res in keys.iter() {
let last_modified: SystemTime = res.updated.clone()
.and_then(|s| DateTime::parse_from_rfc3339(&s).ok())
.map(|s| s.into())
.unwrap_or(SystemTime::now());
let size = res.size.clone().unwrap_or("0".to_string()).parse::<u64>().unwrap();
let key = res.name.clone();
result.keys.push(
ListingObject{
key: self.gcs_object_to_relative_path(&key),
last_modified,
size,
}
);
if let Some(mut mk) = max_keys {
assert!(mk > 0);
mk -= 1;
if mk == 0 {
tracing::debug!("reached limit set by max_keys");
yield Ok(result);
break 'outer;
}
max_keys = Some(mk);
};
}
result.prefixes.extend(prefixes.iter().filter_map(|p| {
Some(
self.gcs_object_to_relative_path(
p.trim_end_matches(REMOTE_STORAGE_PREFIX_SEPARATOR)
),
)
}));
yield Ok(result);
continuation_token = match resp.next_page_token {
Some(token) => {
list_uri = list_uri + "&pageToken=" + &token;
Some(token)
},
None => break
}
}
}
}
async fn head_object(
&self,
key: &RemotePath,
cancel: &CancellationToken,
) -> Result<ListingObject, DownloadError> {
let kind = RequestKind::Head;
todo!();
}
async fn upload(
&self,
from: impl Stream<Item = std::io::Result<Bytes>> + Send + Sync + 'static,
from_size_bytes: usize,
to: &RemotePath,
metadata: Option<StorageMetadata>,
cancel: &CancellationToken,
) -> anyhow::Result<()> {
let kind = RequestKind::Put;
let _permit = self.permit(kind, cancel).await?;
let started_at = start_measuring_requests(kind);
let upload = self.put_object(from, from_size_bytes, to, cancel);
let upload = tokio::time::timeout(self.timeout, upload);
let res = tokio::select! {
res = upload => res,
_ = cancel.cancelled() => return Err(TimeoutOrCancel::Cancel.into()),
};
if let Ok(inner) = &res {
// do not incl. timeouts as errors in metrics but cancellations
let started_at = ScopeGuard::into_inner(started_at);
crate::metrics::BUCKET_METRICS
.req_seconds
.observe_elapsed(kind, inner, started_at);
}
match res {
Ok(Ok(_put)) => Ok(()),
Ok(Err(sdk)) => Err(sdk.into()),
Err(_timeout) => Err(TimeoutOrCancel::Timeout.into()),
}
}
async fn copy(
&self,
from: &RemotePath,
to: &RemotePath,
cancel: &CancellationToken,
) -> anyhow::Result<()> {
let kind = RequestKind::Copy;
let _permit = self.permit(kind, cancel).await?;
let timeout = tokio::time::sleep(self.timeout);
let started_at = start_measuring_requests(kind);
// we need to specify bucket_name as a prefix
let copy_source = format!(
"{}/{}",
self.bucket_name,
self.relative_path_to_gcs_object(from)
);
todo!();
}
async fn download(
&self,
from: &RemotePath,
opts: &DownloadOpts,
cancel: &CancellationToken,
) -> Result<Download, DownloadError> {
// if prefix is not none then download file `prefix/from`
// if prefix is none then download file `from`
self.get_object(
GetObjectRequest {
bucket: self.bucket_name.clone(),
key: self
.relative_path_to_gcs_object(from)
.trim_start_matches("/")
.to_string(),
etag: opts.etag.as_ref().map(|e| e.to_string()),
range: opts.byte_range_header(),
},
cancel,
)
.await
}
async fn delete_objects(
&self,
paths: &[RemotePath],
cancel: &CancellationToken,
) -> anyhow::Result<()> {
let kind = RequestKind::Delete;
let permit = self.permit(kind, cancel).await?;
let mut delete_objects: Vec<String> = Vec::with_capacity(paths.len());
let delete_objects: Vec<String> = paths
.iter()
.map(|i| self.relative_path_to_gcs_object(i))
.collect();
self.delete_oids(&delete_objects, cancel, &permit).await
}
fn max_keys_per_delete(&self) -> usize {
MAX_KEYS_PER_DELETE_GCS
}
async fn delete(&self, path: &RemotePath, cancel: &CancellationToken) -> anyhow::Result<()> {
let paths = std::array::from_ref(path);
self.delete_objects(paths, cancel).await
}
async fn time_travel_recover(
&self,
prefix: Option<&RemotePath>,
timestamp: SystemTime,
done_if_after: SystemTime,
cancel: &CancellationToken,
) -> Result<(), TimeTravelError> {
Ok(())
}
}
// ---------
#[derive(Serialize, Deserialize, Debug)]
#[serde(rename_all = "snake_case")]
pub struct GCSListResponse {
#[serde(rename = "nextPageToken")]
pub next_page_token: Option<String>,
pub items: Option<Vec<GCSObject>>,
pub prefixes: Option<Vec<String>>,
}
#[derive(Serialize, Deserialize, Debug)]
#[serde(rename_all = "snake_case")]
pub struct GCSObject {
pub name: String,
pub bucket: String,
pub generation: String,
pub metageneration: String,
#[serde(rename = "contentType")]
pub content_type: Option<String>,
#[serde(rename = "storageClass")]
pub storage_class: String,
pub size: Option<String>,
#[serde(rename = "md5Hash")]
pub md5_hash: Option<String>,
pub crc32c: String,
pub etag: Option<String>,
#[serde(rename = "timeCreated")]
pub time_created: String,
pub updated: Option<String>,
#[serde(rename = "timeStorageClassUpdated")]
pub time_storage_class_updated: String,
#[serde(rename = "timeFinalized")]
pub time_finalized: String,
pub metadata: Option<HashMap<String, String>>,
}
impl GCSListResponse {
pub fn contents(&self) -> &[GCSObject] {
self.items.as_deref().unwrap_or_default()
}
pub fn common_prefixes(&self) -> &[String] {
self.prefixes.as_deref().unwrap_or_default()
}
}
#[cfg(test)]
mod tests {
use super::*;
use gcp_auth;
use std::num::NonZero;
use std::pin::pin;
use std::sync::Arc;
const BUFFER_SIZE: usize = 32 * 1024;
// TODO what does Neon want here for integration tests?
const BUCKET: &str = "https://storage.googleapis.com/storage/v1/b/my-test-bucket";
#[tokio::test]
async fn list_returns_keys_from_bucket() {
let provider = gcp_auth::provider().await.unwrap();
let gcs = GCSBucket {
token_provider: Arc::clone(&provider),
bucket_name: BUCKET.to_string(),
prefix_in_bucket: None,
max_keys_per_list_response: Some(100),
concurrency_limiter: ConcurrencyLimiter::new(100),
timeout: std::time::Duration::from_secs(120),
};
let cancel = CancellationToken::new();
let remote_prefix = "box/tiff/2023/TN".to_string();
let max_keys: u32 = 100;
let mut stream = pin!(gcs.list_streaming(Some(remote_prefix), NonZero::new(max_keys)));
let mut combined = stream
.next()
.await
.expect("At least one item required")
.unwrap();
while let Some(list) = stream.next().await {
let list = list.unwrap();
combined.keys.extend(list.keys.into_iter());
combined.prefixes.extend_from_slice(&list.prefixes);
}
for key in combined.keys.iter() {
println!("Item: {} -- {:?}", key.key, key.last_modified);
}
assert_ne!(0, combined.keys.len());
}
}

View File

@@ -12,6 +12,7 @@
mod azure_blob;
mod config;
mod error;
mod gcs_bucket;
mod local_fs;
mod metrics;
mod s3_bucket;
@@ -42,6 +43,7 @@ use tokio_util::sync::CancellationToken;
use tracing::info;
pub use self::azure_blob::AzureBlobStorage;
pub use self::gcs_bucket::GCSBucket;
pub use self::local_fs::LocalFs;
pub use self::s3_bucket::S3Bucket;
pub use self::simulate_failures::UnreliableWrapper;
@@ -80,8 +82,12 @@ pub const MAX_KEYS_PER_DELETE_S3: usize = 1000;
/// <https://learn.microsoft.com/en-us/rest/api/storageservices/blob-batch>
pub const MAX_KEYS_PER_DELETE_AZURE: usize = 256;
pub const MAX_KEYS_PER_DELETE_GCS: usize = 1000;
const REMOTE_STORAGE_PREFIX_SEPARATOR: char = '/';
const GCS_SCOPES: &[&str] = &["https://www.googleapis.com/auth/cloud-platform"];
/// Path on the remote storage, relative to some inner prefix.
/// The prefix is an implementation detail, that allows representing local paths
/// as the remote ones, stripping the local storage prefix away.
@@ -439,6 +445,7 @@ pub enum GenericRemoteStorage<Other: Clone = Arc<UnreliableWrapper>> {
AwsS3(Arc<S3Bucket>),
AzureBlob(Arc<AzureBlobStorage>),
Unreliable(Other),
GCS(Arc<GCSBucket>),
}
impl<Other: RemoteStorage> GenericRemoteStorage<Arc<Other>> {
@@ -455,6 +462,7 @@ impl<Other: RemoteStorage> GenericRemoteStorage<Arc<Other>> {
Self::AwsS3(s) => s.list(prefix, mode, max_keys, cancel).await,
Self::AzureBlob(s) => s.list(prefix, mode, max_keys, cancel).await,
Self::Unreliable(s) => s.list(prefix, mode, max_keys, cancel).await,
Self::GCS(s) => s.list(prefix, mode, max_keys, cancel).await,
}
}
@@ -472,6 +480,7 @@ impl<Other: RemoteStorage> GenericRemoteStorage<Arc<Other>> {
Self::AwsS3(s) => Box::pin(s.list_streaming(prefix, mode, max_keys, cancel)),
Self::AzureBlob(s) => Box::pin(s.list_streaming(prefix, mode, max_keys, cancel)),
Self::Unreliable(s) => Box::pin(s.list_streaming(prefix, mode, max_keys, cancel)),
Self::GCS(s) => Box::pin(s.list_streaming(prefix, mode, max_keys, cancel)),
}
}
@@ -486,6 +495,7 @@ impl<Other: RemoteStorage> GenericRemoteStorage<Arc<Other>> {
Self::AwsS3(s) => s.head_object(key, cancel).await,
Self::AzureBlob(s) => s.head_object(key, cancel).await,
Self::Unreliable(s) => s.head_object(key, cancel).await,
Self::GCS(_) => todo!(),
}
}
@@ -503,6 +513,7 @@ impl<Other: RemoteStorage> GenericRemoteStorage<Arc<Other>> {
Self::AwsS3(s) => s.upload(from, data_size_bytes, to, metadata, cancel).await,
Self::AzureBlob(s) => s.upload(from, data_size_bytes, to, metadata, cancel).await,
Self::Unreliable(s) => s.upload(from, data_size_bytes, to, metadata, cancel).await,
Self::GCS(s) => s.upload(from, data_size_bytes, to, metadata, cancel).await,
}
}
@@ -518,6 +529,7 @@ impl<Other: RemoteStorage> GenericRemoteStorage<Arc<Other>> {
Self::AwsS3(s) => s.download(from, opts, cancel).await,
Self::AzureBlob(s) => s.download(from, opts, cancel).await,
Self::Unreliable(s) => s.download(from, opts, cancel).await,
Self::GCS(s) => s.download(from, opts, cancel).await,
}
}
@@ -532,6 +544,7 @@ impl<Other: RemoteStorage> GenericRemoteStorage<Arc<Other>> {
Self::AwsS3(s) => s.delete(path, cancel).await,
Self::AzureBlob(s) => s.delete(path, cancel).await,
Self::Unreliable(s) => s.delete(path, cancel).await,
Self::GCS(s) => s.delete(path, cancel).await,
}
}
@@ -546,6 +559,7 @@ impl<Other: RemoteStorage> GenericRemoteStorage<Arc<Other>> {
Self::AwsS3(s) => s.delete_objects(paths, cancel).await,
Self::AzureBlob(s) => s.delete_objects(paths, cancel).await,
Self::Unreliable(s) => s.delete_objects(paths, cancel).await,
Self::GCS(s) => s.delete_objects(paths, cancel).await,
}
}
@@ -556,6 +570,7 @@ impl<Other: RemoteStorage> GenericRemoteStorage<Arc<Other>> {
Self::AwsS3(s) => s.max_keys_per_delete(),
Self::AzureBlob(s) => s.max_keys_per_delete(),
Self::Unreliable(s) => s.max_keys_per_delete(),
Self::GCS(s) => s.max_keys_per_delete(),
}
}
@@ -570,6 +585,7 @@ impl<Other: RemoteStorage> GenericRemoteStorage<Arc<Other>> {
Self::AwsS3(s) => s.delete_prefix(prefix, cancel).await,
Self::AzureBlob(s) => s.delete_prefix(prefix, cancel).await,
Self::Unreliable(s) => s.delete_prefix(prefix, cancel).await,
Self::GCS(_) => todo!(),
}
}
@@ -585,6 +601,7 @@ impl<Other: RemoteStorage> GenericRemoteStorage<Arc<Other>> {
Self::AwsS3(s) => s.copy(from, to, cancel).await,
Self::AzureBlob(s) => s.copy(from, to, cancel).await,
Self::Unreliable(s) => s.copy(from, to, cancel).await,
Self::GCS(_) => todo!(),
}
}
@@ -613,17 +630,25 @@ impl<Other: RemoteStorage> GenericRemoteStorage<Arc<Other>> {
s.time_travel_recover(prefix, timestamp, done_if_after, cancel)
.await
}
Self::GCS(_) => todo!(),
}
}
}
impl GenericRemoteStorage {
pub async fn from_config(storage_config: &RemoteStorageConfig) -> anyhow::Result<Self> {
info!("RemoteStorageConfig: {:?}", storage_config);
let timeout = storage_config.timeout;
// If somkeone overrides timeout to be small without adjusting small_timeout, then adjust it automatically
// If someone overrides timeout to be small without adjusting small_timeout, then adjust it automatically
let small_timeout = std::cmp::min(storage_config.small_timeout, timeout);
info!(
"RemoteStorageConfig's storage attribute: {:?}",
storage_config.storage
);
Ok(match &storage_config.storage {
RemoteStorageKind::LocalFs { local_path: path } => {
info!("Using fs root '{path}' as a remote storage");
@@ -661,6 +686,16 @@ impl GenericRemoteStorage {
small_timeout,
)?))
}
RemoteStorageKind::GCS(gcs_config) => {
let google_application_credentials =
std::env::var("GOOGLE_APPLICATION_CREDENTIALS")
.unwrap_or_else(|_| "<none>".into());
info!(
"Using gcs bucket '{}' as a remote storage, prefix in bucket: '{:?}', GOOGLE_APPLICATION_CREDENTIALS: {google_application_credentials }",
gcs_config.bucket_name, gcs_config.prefix_in_bucket
);
Self::GCS(Arc::new(GCSBucket::new(gcs_config, timeout).await?))
}
})
}
@@ -690,6 +725,7 @@ impl GenericRemoteStorage {
Self::AwsS3(s) => Some(s.bucket_name()),
Self::AzureBlob(s) => Some(s.container_name()),
Self::Unreliable(_s) => None,
Self::GCS(s) => Some(s.bucket_name()),
}
}
}

View File

@@ -50,6 +50,7 @@ impl UnreliableWrapper {
GenericRemoteStorage::Unreliable(_s) => {
panic!("Can't wrap unreliable wrapper unreliably")
}
GenericRemoteStorage::GCS(_) => todo!(),
};
UnreliableWrapper {
inner,

View File

@@ -1,5 +1,5 @@
[package]
name = "endpoint_storage"
name = "object_storage"
version = "0.0.1"
edition.workspace = true
license.workspace = true

View File

@@ -2,7 +2,7 @@ use anyhow::anyhow;
use axum::body::{Body, Bytes};
use axum::response::{IntoResponse, Response};
use axum::{Router, http::StatusCode};
use endpoint_storage::{PrefixS3Path, S3Path, Storage, bad_request, internal_error, not_found, ok};
use object_storage::{PrefixS3Path, S3Path, Storage, bad_request, internal_error, not_found, ok};
use remote_storage::TimeoutOrCancel;
use remote_storage::{DownloadError, DownloadOpts, GenericRemoteStorage, RemotePath};
use std::{sync::Arc, time::SystemTime, time::UNIX_EPOCH};
@@ -46,12 +46,12 @@ async fn metrics() -> Result {
async fn get(S3Path { path }: S3Path, state: State) -> Result {
info!(%path, "downloading");
let download_err = |err| {
if let DownloadError::NotFound = err {
info!(%path, %err, "downloading"); // 404 is not an issue of _this_ service
let download_err = |e| {
if let DownloadError::NotFound = e {
info!(%path, %e, "downloading"); // 404 is not an issue of _this_ service
return not_found(&path);
}
internal_error(err, &path, "downloading")
internal_error(e, &path, "downloading")
};
let cancel = state.cancel.clone();
let opts = &DownloadOpts::default();
@@ -249,7 +249,7 @@ mod tests {
};
let proxy = Storage {
auth: endpoint_storage::JwtAuth::new(TEST_PUB_KEY_ED25519).unwrap(),
auth: object_storage::JwtAuth::new(TEST_PUB_KEY_ED25519).unwrap(),
storage,
cancel: cancel.clone(),
max_upload_file_limit: usize::MAX,
@@ -343,14 +343,14 @@ MC4CAQAwBQYDK2VwBCIEID/Drmc1AA6U/znNRWpF3zEGegOATQxfkdWxitcOMsIH
TimelineId::from_array([0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 1, 2, 3, 4, 5, 7]);
const ENDPOINT_ID: &str = "ep-winter-frost-a662z3vg";
fn token() -> String {
let claims = endpoint_storage::Claims {
let claims = object_storage::Claims {
tenant_id: TENANT_ID,
timeline_id: TIMELINE_ID,
endpoint_id: ENDPOINT_ID.into(),
exp: u64::MAX,
};
let key = jsonwebtoken::EncodingKey::from_ed_pem(TEST_PRIV_KEY_ED25519).unwrap();
let header = jsonwebtoken::Header::new(endpoint_storage::VALIDATION_ALGO);
let header = jsonwebtoken::Header::new(object_storage::VALIDATION_ALGO);
jsonwebtoken::encode(&header, &claims, &key).unwrap()
}
@@ -364,10 +364,7 @@ MC4CAQAwBQYDK2VwBCIEID/Drmc1AA6U/znNRWpF3zEGegOATQxfkdWxitcOMsIH
vec![TIMELINE_ID.to_string(), TimelineId::generate().to_string()],
vec![ENDPOINT_ID, "ep-ololo"]
)
// first one is fully valid path, second path is valid for GET as
// read paths may have different endpoint if tenant and timeline matches
// (needed for prewarming RO->RW replica)
.skip(2);
.skip(1);
for ((uri, method), (tenant, timeline, endpoint)) in iproduct!(routes(), args) {
info!(%uri, %method, %tenant, %timeline, %endpoint);
@@ -478,16 +475,6 @@ MC4CAQAwBQYDK2VwBCIEID/Drmc1AA6U/znNRWpF3zEGegOATQxfkdWxitcOMsIH
requests_chain(chain.into_iter(), |_| token()).await;
}
#[testlog(tokio::test)]
async fn read_other_endpoint_data() {
let uri = format!("/{TENANT_ID}/{TIMELINE_ID}/other_endpoint/key");
let chain = vec![
(uri.clone(), "GET", "", StatusCode::NOT_FOUND, false),
(uri.clone(), "PUT", "", StatusCode::UNAUTHORIZED, false),
];
requests_chain(chain.into_iter(), |_| token()).await;
}
fn delete_prefix_token(uri: &str) -> String {
use serde::Serialize;
let parts = uri.split("/").collect::<Vec<&str>>();
@@ -495,7 +482,7 @@ MC4CAQAwBQYDK2VwBCIEID/Drmc1AA6U/znNRWpF3zEGegOATQxfkdWxitcOMsIH
struct PrefixClaims {
tenant_id: TenantId,
timeline_id: Option<TimelineId>,
endpoint_id: Option<endpoint_storage::EndpointId>,
endpoint_id: Option<object_storage::EndpointId>,
exp: u64,
}
let claims = PrefixClaims {
@@ -505,7 +492,7 @@ MC4CAQAwBQYDK2VwBCIEID/Drmc1AA6U/znNRWpF3zEGegOATQxfkdWxitcOMsIH
exp: u64::MAX,
};
let key = jsonwebtoken::EncodingKey::from_ed_pem(TEST_PRIV_KEY_ED25519).unwrap();
let header = jsonwebtoken::Header::new(endpoint_storage::VALIDATION_ALGO);
let header = jsonwebtoken::Header::new(object_storage::VALIDATION_ALGO);
jsonwebtoken::encode(&header, &claims, &key).unwrap()
}

View File

@@ -169,19 +169,10 @@ impl FromRequestParts<Arc<Storage>> for S3Path {
.auth
.decode(bearer.token())
.map_err(|e| bad_request(e, "decoding token"))?;
// Read paths may have different endpoint ids. For readonly -> readwrite replica
// prewarming, endpoint must read other endpoint's data.
let endpoint_id = if parts.method == axum::http::Method::GET {
claims.endpoint_id.clone()
} else {
path.endpoint_id.clone()
};
let route = Claims {
tenant_id: path.tenant_id,
timeline_id: path.timeline_id,
endpoint_id,
endpoint_id: path.endpoint_id.clone(),
exp: claims.exp,
};
if route != claims {

View File

@@ -1,4 +1,4 @@
//! `endpoint_storage` is a service which provides API for uploading and downloading
//! `object_storage` is a service which provides API for uploading and downloading
//! files. It is used by compute and control plane for accessing LFC prewarm data.
//! This service is deployed either as a separate component or as part of compute image
//! for large computes.
@@ -33,7 +33,7 @@ async fn main() -> anyhow::Result<()> {
let config: String = std::env::args().skip(1).take(1).collect();
if config.is_empty() {
anyhow::bail!("Usage: endpoint_storage config.json")
anyhow::bail!("Usage: object_storage config.json")
}
info!("Reading config from {config}");
let config = std::fs::read_to_string(config.clone())?;
@@ -41,7 +41,7 @@ async fn main() -> anyhow::Result<()> {
info!("Reading pemfile from {}", config.pemfile.clone());
let pemfile = std::fs::read(config.pemfile.clone())?;
info!("Loading public key from {}", config.pemfile.clone());
let auth = endpoint_storage::JwtAuth::new(&pemfile)?;
let auth = object_storage::JwtAuth::new(&pemfile)?;
let listener = tokio::net::TcpListener::bind(config.listen).await.unwrap();
info!("listening on {}", listener.local_addr().unwrap());
@@ -50,7 +50,7 @@ async fn main() -> anyhow::Result<()> {
let cancel = tokio_util::sync::CancellationToken::new();
app::check_storage_permissions(&storage, cancel.clone()).await?;
let proxy = std::sync::Arc::new(endpoint_storage::Storage {
let proxy = std::sync::Arc::new(object_storage::Storage {
auth,
storage,
cancel: cancel.clone(),

View File

@@ -263,9 +263,7 @@ where
while let Some((tenant_id, tenant)) = tenants.next().await {
let mut tenant_resident_size = 0;
let timelines = tenant.list_timelines();
let timelines_len = timelines.len();
for timeline in timelines {
for timeline in tenant.list_timelines() {
let timeline_id = timeline.timeline_id;
match TimelineSnapshot::collect(&timeline, ctx) {
@@ -291,11 +289,6 @@ where
tenant_resident_size += timeline.resident_physical_size();
}
if timelines_len == 0 {
// Force set it to 1 byte to avoid not being reported -- all timelines are offloaded.
tenant_resident_size = 1;
}
let snap = TenantSnapshot::collect(&tenant, tenant_resident_size);
snap.to_metrics(tenant_id, Utc::now(), cache, &mut current_metrics);
}

View File

@@ -1285,10 +1285,6 @@ impl Timeline {
reconstruct_state: &mut ValuesReconstructState,
ctx: &RequestContext,
) -> Result<BTreeMap<Key, Result<Bytes, PageReconstructError>>, GetVectoredError> {
if query.is_empty() {
return Ok(BTreeMap::default());
}
let read_path = if self.conf.enable_read_path_debugging || ctx.read_path_debug() {
Some(ReadPath::new(
query.total_keyspace(),

View File

@@ -803,13 +803,7 @@ neon_create(SMgrRelation reln, ForkNumber forkNum, bool isRedo)
case RELPERSISTENCE_TEMP:
case RELPERSISTENCE_UNLOGGED:
#ifdef DEBUG_COMPARE_LOCAL
mdcreate(reln, forkNum, forkNum == INIT_FORKNUM || isRedo);
if (forkNum == MAIN_FORKNUM)
mdcreate(reln, INIT_FORKNUM, true);
#else
mdcreate(reln, forkNum, isRedo);
#endif
return;
default:
@@ -1979,10 +1973,6 @@ neon_start_unlogged_build(SMgrRelation reln)
case RELPERSISTENCE_UNLOGGED:
unlogged_build_rel = reln;
unlogged_build_phase = UNLOGGED_BUILD_NOT_PERMANENT;
#ifdef DEBUG_COMPARE_LOCAL
if (!IsParallelWorker())
mdcreate(reln, INIT_FORKNUM, true);
#endif
return;
default:
@@ -2005,14 +1995,12 @@ neon_start_unlogged_build(SMgrRelation reln)
* FIXME: should we pass isRedo true to create the tablespace dir if it
* doesn't exist? Is it needed?
*/
if (!IsParallelWorker())
{
#ifndef DEBUG_COMPARE_LOCAL
if (!IsParallelWorker())
mdcreate(reln, MAIN_FORKNUM, false);
#else
mdcreate(reln, INIT_FORKNUM, true);
mdcreate(reln, INIT_FORKNUM, false);
#endif
}
}
/*
@@ -2111,12 +2099,12 @@ neon_end_unlogged_build(SMgrRelation reln)
#ifndef DEBUG_COMPARE_LOCAL
/* use isRedo == true, so that we drop it immediately */
mdunlink(rinfob, forknum, true);
#else
mdunlink(rinfob, INIT_FORKNUM, true);
#endif
}
#ifdef DEBUG_COMPARE_LOCAL
mdunlink(rinfob, INIT_FORKNUM, true);
#endif
}
unlogged_build_rel = NULL;
unlogged_build_phase = UNLOGGED_BUILD_NOT_IN_PROGRESS;
}

View File

@@ -91,7 +91,6 @@ mod jemalloc;
mod logging;
mod metrics;
mod parse;
mod pglb;
mod protocol2;
mod proxy;
mod rate_limiter;

View File

@@ -1,193 +0,0 @@
#![allow(dead_code, reason = "TODO: work in progress")]
use std::pin::{Pin, pin};
use std::sync::Arc;
use std::sync::atomic::{AtomicUsize, Ordering};
use std::task::{Context, Poll};
use std::{fmt, io};
use tokio::io::{AsyncRead, AsyncWrite, DuplexStream, ReadBuf};
use tokio::sync::mpsc;
const STREAM_CHANNEL_SIZE: usize = 16;
const MAX_STREAM_BUFFER_SIZE: usize = 4096;
#[derive(Debug)]
pub struct Connection {
stream_sender: mpsc::Sender<Stream>,
stream_receiver: mpsc::Receiver<Stream>,
stream_id_counter: Arc<AtomicUsize>,
}
impl Connection {
pub fn new() -> (Connection, Connection) {
let (sender_a, receiver_a) = mpsc::channel(STREAM_CHANNEL_SIZE);
let (sender_b, receiver_b) = mpsc::channel(STREAM_CHANNEL_SIZE);
let stream_id_counter = Arc::new(AtomicUsize::new(1));
let conn_a = Connection {
stream_sender: sender_a,
stream_receiver: receiver_b,
stream_id_counter: Arc::clone(&stream_id_counter),
};
let conn_b = Connection {
stream_sender: sender_b,
stream_receiver: receiver_a,
stream_id_counter,
};
(conn_a, conn_b)
}
#[inline]
fn next_stream_id(&self) -> StreamId {
StreamId(self.stream_id_counter.fetch_add(1, Ordering::Relaxed))
}
#[tracing::instrument(skip_all, fields(stream_id = tracing::field::Empty, err))]
pub async fn open_stream(&self) -> io::Result<Stream> {
let (local, remote) = tokio::io::duplex(MAX_STREAM_BUFFER_SIZE);
let stream_id = self.next_stream_id();
tracing::Span::current().record("stream_id", stream_id.0);
let local = Stream {
inner: local,
id: stream_id,
};
let remote = Stream {
inner: remote,
id: stream_id,
};
self.stream_sender
.send(remote)
.await
.map_err(io::Error::other)?;
Ok(local)
}
#[tracing::instrument(skip_all, fields(stream_id = tracing::field::Empty, err))]
pub async fn accept_stream(&mut self) -> io::Result<Option<Stream>> {
Ok(self.stream_receiver.recv().await.inspect(|stream| {
tracing::Span::current().record("stream_id", stream.id.0);
}))
}
}
#[derive(Copy, Clone, Debug)]
pub struct StreamId(usize);
impl fmt::Display for StreamId {
#[inline]
fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result {
write!(f, "{}", self.0)
}
}
// TODO: Proper closing. Currently Streams can outlive their Connections.
// Carry WeakSender and check strong_count?
#[derive(Debug)]
pub struct Stream {
inner: DuplexStream,
id: StreamId,
}
impl Stream {
#[inline]
pub fn id(&self) -> StreamId {
self.id
}
}
impl AsyncRead for Stream {
#[tracing::instrument(level = "debug", skip_all, fields(stream_id = %self.id))]
#[inline]
fn poll_read(
mut self: Pin<&mut Self>,
cx: &mut Context<'_>,
buf: &mut ReadBuf<'_>,
) -> Poll<io::Result<()>> {
pin!(&mut self.inner).poll_read(cx, buf)
}
}
impl AsyncWrite for Stream {
#[tracing::instrument(level = "debug", skip_all, fields(stream_id = %self.id))]
#[inline]
fn poll_write(
mut self: Pin<&mut Self>,
cx: &mut Context<'_>,
buf: &[u8],
) -> Poll<Result<usize, io::Error>> {
pin!(&mut self.inner).poll_write(cx, buf)
}
#[tracing::instrument(level = "debug", skip_all, fields(stream_id = %self.id))]
#[inline]
fn poll_flush(mut self: Pin<&mut Self>, cx: &mut Context<'_>) -> Poll<Result<(), io::Error>> {
pin!(&mut self.inner).poll_flush(cx)
}
#[tracing::instrument(level = "debug", skip_all, fields(stream_id = %self.id))]
#[inline]
fn poll_shutdown(
mut self: Pin<&mut Self>,
cx: &mut Context<'_>,
) -> Poll<Result<(), io::Error>> {
pin!(&mut self.inner).poll_shutdown(cx)
}
#[tracing::instrument(level = "debug", skip_all, fields(stream_id = %self.id))]
#[inline]
fn poll_write_vectored(
mut self: Pin<&mut Self>,
cx: &mut Context<'_>,
bufs: &[io::IoSlice<'_>],
) -> Poll<Result<usize, io::Error>> {
pin!(&mut self.inner).poll_write_vectored(cx, bufs)
}
#[inline]
fn is_write_vectored(&self) -> bool {
self.inner.is_write_vectored()
}
}
#[cfg(test)]
mod tests {
use tokio::io::{AsyncReadExt, AsyncWriteExt};
use super::*;
#[tokio::test]
async fn test_simple_roundtrip() {
let (client, mut server) = Connection::new();
let server_task = tokio::spawn(async move {
while let Some(mut stream) = server.accept_stream().await.unwrap() {
tokio::spawn(async move {
let mut buf = [0; 64];
loop {
match stream.read(&mut buf).await.unwrap() {
0 => break,
n => stream.write(&buf[..n]).await.unwrap(),
};
}
});
}
});
let mut stream = client.open_stream().await.unwrap();
stream.write_all(b"hello!").await.unwrap();
let mut buf = [0; 64];
let n = stream.read(&mut buf).await.unwrap();
assert_eq!(n, 6);
assert_eq!(&buf[..n], b"hello!");
drop(stream);
drop(client);
server_task.await.unwrap();
}
}

View File

@@ -1 +0,0 @@
pub mod inprocess;

View File

@@ -14,7 +14,6 @@ use clap::{ArgAction, Parser};
use futures::future::BoxFuture;
use futures::stream::FuturesUnordered;
use futures::{FutureExt, StreamExt};
use http_utils::tls_certs::ReloadingCertificateResolver;
use metrics::set_build_info_metric;
use remote_storage::RemoteStorageConfig;
use safekeeper::defaults::{
@@ -24,8 +23,8 @@ use safekeeper::defaults::{
DEFAULT_SSL_CERT_RELOAD_PERIOD, DEFAULT_SSL_KEY_FILE,
};
use safekeeper::{
BACKGROUND_RUNTIME, BROKER_RUNTIME, GlobalTimelines, HTTP_RUNTIME, SafeKeeperConf,
WAL_SERVICE_RUNTIME, broker, control_file, http, wal_backup, wal_service,
BROKER_RUNTIME, GlobalTimelines, HTTP_RUNTIME, SafeKeeperConf, WAL_SERVICE_RUNTIME, broker,
control_file, http, wal_backup, wal_service,
};
use sd_notify::NotifyState;
use storage_broker::{DEFAULT_ENDPOINT, Uri};
@@ -216,21 +215,16 @@ struct Args {
ssl_cert_file: Utf8PathBuf,
/// Period to reload certificate and private key from files.
#[arg(long, value_parser = humantime::parse_duration, default_value = DEFAULT_SSL_CERT_RELOAD_PERIOD)]
ssl_cert_reload_period: Duration,
pub ssl_cert_reload_period: Duration,
/// Trusted root CA certificates to use in https APIs.
#[arg(long)]
ssl_ca_file: Option<Utf8PathBuf>,
pub ssl_ca_file: Option<Utf8PathBuf>,
/// Flag to use https for requests to peer's safekeeper API.
#[arg(long)]
use_https_safekeeper_api: bool,
pub use_https_safekeeper_api: bool,
/// Path to the JWT auth token used to authenticate with other safekeepers.
#[arg(long)]
auth_token_path: Option<Utf8PathBuf>,
/// Enable TLS in WAL service API.
/// Does not force TLS: the client negotiates TLS usage during the handshake.
/// Uses key and certificate from ssl_key_file/ssl_cert_file.
#[arg(long)]
enable_tls_wal_service_api: bool,
}
// Like PathBufValueParser, but allows empty string.
@@ -424,7 +418,6 @@ async fn main() -> anyhow::Result<()> {
ssl_cert_reload_period: args.ssl_cert_reload_period,
ssl_ca_certs,
use_https_safekeeper_api: args.use_https_safekeeper_api,
enable_tls_wal_service_api: args.enable_tls_wal_service_api,
});
// initialize sentry if SENTRY_DSN is provided
@@ -524,36 +517,6 @@ async fn start_safekeeper(conf: Arc<SafeKeeperConf>) -> Result<()> {
info!("running in current thread runtime");
}
let tls_server_config = if conf.listen_https_addr.is_some() || conf.enable_tls_wal_service_api {
let ssl_key_file = conf.ssl_key_file.clone();
let ssl_cert_file = conf.ssl_cert_file.clone();
let ssl_cert_reload_period = conf.ssl_cert_reload_period;
// Create resolver in BACKGROUND_RUNTIME, so the background certificate reloading
// task is run in this runtime.
let cert_resolver = current_thread_rt
.as_ref()
.unwrap_or_else(|| BACKGROUND_RUNTIME.handle())
.spawn(async move {
ReloadingCertificateResolver::new(
"main",
&ssl_key_file,
&ssl_cert_file,
ssl_cert_reload_period,
)
.await
})
.await??;
let config = rustls::ServerConfig::builder()
.with_no_client_auth()
.with_cert_resolver(cert_resolver);
Some(Arc::new(config))
} else {
None
};
let wal_service_handle = current_thread_rt
.as_ref()
.unwrap_or_else(|| WAL_SERVICE_RUNTIME.handle())
@@ -561,9 +524,6 @@ async fn start_safekeeper(conf: Arc<SafeKeeperConf>) -> Result<()> {
conf.clone(),
pg_listener,
Scope::SafekeeperData,
conf.enable_tls_wal_service_api
.then(|| tls_server_config.clone())
.flatten(),
global_timelines.clone(),
))
// wrap with task name for error reporting
@@ -592,9 +552,6 @@ async fn start_safekeeper(conf: Arc<SafeKeeperConf>) -> Result<()> {
conf.clone(),
pg_listener_tenant_only,
Scope::Tenant,
conf.enable_tls_wal_service_api
.then(|| tls_server_config.clone())
.flatten(),
global_timelines.clone(),
))
// wrap with task name for error reporting
@@ -620,7 +577,6 @@ async fn start_safekeeper(conf: Arc<SafeKeeperConf>) -> Result<()> {
.spawn(http::task_main_https(
conf.clone(),
https_listener,
tls_server_config.expect("tls_server_config is set earlier if https is enabled"),
global_timelines.clone(),
))
.map(|res| ("HTTPS service main".to_owned(), res));

View File

@@ -1,6 +1,7 @@
pub mod routes;
use std::sync::Arc;
use http_utils::tls_certs::ReloadingCertificateResolver;
pub use routes::make_router;
pub use safekeeper_api::models;
use tokio_util::sync::CancellationToken;
@@ -27,10 +28,21 @@ pub async fn task_main_http(
pub async fn task_main_https(
conf: Arc<SafeKeeperConf>,
https_listener: std::net::TcpListener,
tls_config: Arc<rustls::ServerConfig>,
global_timelines: Arc<GlobalTimelines>,
) -> anyhow::Result<()> {
let tls_acceptor = tokio_rustls::TlsAcceptor::from(tls_config);
let cert_resolver = ReloadingCertificateResolver::new(
"main",
&conf.ssl_key_file,
&conf.ssl_cert_file,
conf.ssl_cert_reload_period,
)
.await?;
let server_config = rustls::ServerConfig::builder()
.with_no_client_auth()
.with_cert_resolver(cert_resolver);
let tls_acceptor = tokio_rustls::TlsAcceptor::from(Arc::new(server_config));
let router = make_router(conf, global_timelines)
.build()

View File

@@ -122,7 +122,6 @@ pub struct SafeKeeperConf {
pub ssl_cert_reload_period: Duration,
pub ssl_ca_certs: Vec<Pem>,
pub use_https_safekeeper_api: bool,
pub enable_tls_wal_service_api: bool,
}
impl SafeKeeperConf {
@@ -173,7 +172,6 @@ impl SafeKeeperConf {
ssl_cert_reload_period: Duration::from_secs(60),
ssl_ca_certs: Vec::new(),
use_https_safekeeper_api: false,
enable_tls_wal_service_api: false,
}
}
}
@@ -211,12 +209,3 @@ pub static WAL_BACKUP_RUNTIME: Lazy<Runtime> = Lazy::new(|| {
.build()
.expect("Failed to create WAL backup runtime")
});
pub static BACKGROUND_RUNTIME: Lazy<Runtime> = Lazy::new(|| {
tokio::runtime::Builder::new_multi_thread()
.thread_name("background worker")
.worker_threads(1) // there is only one task now (ssl certificate reloading), having more threads doesn't make sense
.enable_all()
.build()
.expect("Failed to create background runtime")
});

View File

@@ -29,7 +29,6 @@ pub async fn task_main(
conf: Arc<SafeKeeperConf>,
pg_listener: std::net::TcpListener,
allowed_auth_scope: Scope,
tls_config: Option<Arc<rustls::ServerConfig>>,
global_timelines: Arc<GlobalTimelines>,
) -> anyhow::Result<()> {
// Tokio's from_std won't do this for us, per its comment.
@@ -44,10 +43,9 @@ pub async fn task_main(
let conf = conf.clone();
let conn_id = issue_connection_id(&mut connection_count);
let global_timelines = global_timelines.clone();
let tls_config = tls_config.clone();
tokio::spawn(
async move {
if let Err(err) = handle_socket(socket, conf, conn_id, allowed_auth_scope, tls_config, global_timelines).await {
if let Err(err) = handle_socket(socket, conf, conn_id, allowed_auth_scope, global_timelines).await {
error!("connection handler exited: {}", err);
}
}
@@ -63,7 +61,6 @@ async fn handle_socket(
conf: Arc<SafeKeeperConf>,
conn_id: ConnectionId,
allowed_auth_scope: Scope,
tls_config: Option<Arc<rustls::ServerConfig>>,
global_timelines: Arc<GlobalTimelines>,
) -> Result<(), QueryError> {
socket.set_nodelay(true)?;
@@ -113,8 +110,7 @@ async fn handle_socket(
auth_pair,
global_timelines,
);
let pgbackend =
PostgresBackend::new_from_io(socket_fd, socket, peer_addr, auth_type, tls_config)?;
let pgbackend = PostgresBackend::new_from_io(socket_fd, socket, peer_addr, auth_type, None)?;
// libpq protocol between safekeeper and walproposer / pageserver
// We don't use shutdown.
pgbackend

View File

@@ -185,7 +185,6 @@ pub fn run_server(os: NodeOs, disk: Arc<SafekeeperDisk>) -> Result<()> {
ssl_cert_reload_period: Duration::ZERO,
ssl_ca_certs: Vec::new(),
use_https_safekeeper_api: false,
enable_tls_wal_service_api: false,
};
let mut global = GlobalMap::new(disk, conf.clone())?;

View File

@@ -151,39 +151,11 @@ impl Service {
"Got {} non-successful responses from initial creation request of total {total_result_count} responses",
remaining.len()
);
let target_sk_count = timeline_persistence.sk_set.len();
let quorum_size = match target_sk_count {
0 => {
return Err(ApiError::InternalServerError(anyhow::anyhow!(
"timeline configured without any safekeepers",
)));
}
1 | 2 => {
#[cfg(feature = "testing")]
{
// In test settings, it is allowed to have one or two safekeepers
target_sk_count
}
#[cfg(not(feature = "testing"))]
{
// The region is misconfigured: we need at least three safekeepers to be configured
// in order to schedule work to them
tracing::warn!(
"couldn't find at least 3 safekeepers for timeline, found: {:?}",
timeline_persistence.sk_set
);
return Err(ApiError::InternalServerError(anyhow::anyhow!(
"couldn't find at least 3 safekeepers to put timeline to"
)));
}
}
_ => target_sk_count / 2 + 1,
};
let success_count = target_sk_count - remaining.len();
if success_count < quorum_size {
if remaining.len() >= 2 {
// Failure
return Err(ApiError::InternalServerError(anyhow::anyhow!(
"not enough successful reconciliations to reach quorum size: {success_count} of {quorum_size} of total {target_sk_count}"
"not enough successful reconciliations to reach quorum, please retry: {} errored",
remaining.len()
)));
}
@@ -520,6 +492,8 @@ impl Service {
pub(crate) async fn safekeepers_for_new_timeline(
&self,
) -> Result<Vec<SafekeeperInfo>, ApiError> {
// Number of safekeepers in different AZs we are looking for
let wanted_count = 3;
let mut all_safekeepers = {
let locked = self.inner.read().unwrap();
locked
@@ -558,19 +532,6 @@ impl Service {
sk.1.id.0,
)
});
// Number of safekeepers in different AZs we are looking for
let wanted_count = match all_safekeepers.len() {
0 => {
return Err(ApiError::InternalServerError(anyhow::anyhow!(
"couldn't find any active safekeeper for new timeline",
)));
}
// Have laxer requirements on testig mode as we don't want to
// spin up three safekeepers for every single test
#[cfg(feature = "testing")]
1 | 2 => all_safekeepers.len(),
_ => 3,
};
let mut sks = Vec::new();
let mut azs = HashSet::new();
for (_sk_util, sk_info, az_id) in all_safekeepers.iter() {

View File

@@ -271,6 +271,7 @@ impl BucketConfig {
"container {}, storage account {:?}, region {}",
config.container_name, config.storage_account, config.container_region
),
RemoteStorageKind::GCS(config) => format!("bucket {}", config.bucket_name),
}
}
pub fn bucket_name(&self) -> Option<&str> {
@@ -418,6 +419,9 @@ async fn init_remote(
config.prefix_in_container.get_or_insert(default_prefix);
}
RemoteStorageKind::LocalFs { .. } => (),
RemoteStorageKind::GCS(config) => {
config.prefix_in_bucket.get_or_insert(default_prefix);
}
}
// We already pass the prefix to the remote client above

View File

@@ -417,14 +417,14 @@ class NeonLocalCli(AbstractNeonCli):
cmd.append(f"--instance-id={instance_id}")
return self.raw_cli(cmd)
def endpoint_storage_start(self, timeout_in_seconds: int | None = None):
cmd = ["endpoint-storage", "start"]
def object_storage_start(self, timeout_in_seconds: int | None = None):
cmd = ["object-storage", "start"]
if timeout_in_seconds is not None:
cmd.append(f"--start-timeout={timeout_in_seconds}s")
return self.raw_cli(cmd)
def endpoint_storage_stop(self, immediate: bool):
cmd = ["endpoint-storage", "stop"]
def object_storage_stop(self, immediate: bool):
cmd = ["object-storage", "stop"]
if immediate:
cmd.extend(["-m", "immediate"])
return self.raw_cli(cmd)

View File

@@ -1029,7 +1029,7 @@ class NeonEnvBuilder:
self.env.broker.assert_no_errors()
self.env.endpoint_storage.assert_no_errors()
self.env.object_storage.assert_no_errors()
try:
self.overlay_cleanup_teardown()
@@ -1126,7 +1126,7 @@ class NeonEnv:
pagectl_env_vars["RUST_LOG"] = self.rust_log_override
self.pagectl = Pagectl(extra_env=pagectl_env_vars, binpath=self.neon_binpath)
self.endpoint_storage = EndpointStorage(self)
self.object_storage = ObjectStorage(self)
# The URL for the pageserver to use as its control_plane_api config
if config.storage_controller_port_override is not None:
@@ -1183,7 +1183,7 @@ class NeonEnv:
},
"safekeepers": [],
"pageservers": [],
"endpoint_storage": {"port": self.port_distributor.get_port()},
"object_storage": {"port": self.port_distributor.get_port()},
"generate_local_ssl_certs": self.generate_local_ssl_certs,
}
@@ -1420,7 +1420,7 @@ class NeonEnv:
self.storage_controller.on_safekeeper_deploy(sk_id, body)
self.storage_controller.safekeeper_scheduling_policy(sk_id, "Active")
self.endpoint_storage.start(timeout_in_seconds=timeout_in_seconds)
self.object_storage.start(timeout_in_seconds=timeout_in_seconds)
def stop(self, immediate=False, ps_assert_metric_no_errors=False, fail_on_endpoint_errors=True):
"""
@@ -1439,7 +1439,7 @@ class NeonEnv:
except Exception as e:
raise_later = e
self.endpoint_storage.stop(immediate=immediate)
self.object_storage.stop(immediate=immediate)
# Stop storage controller before pageservers: we don't want it to spuriously
# detect a pageserver "failure" during test teardown
@@ -2660,24 +2660,24 @@ class NeonStorageController(MetricsGetter, LogUtils):
self.stop(immediate=True)
class EndpointStorage(LogUtils):
class ObjectStorage(LogUtils):
def __init__(self, env: NeonEnv):
service_dir = env.repo_dir / "endpoint_storage"
super().__init__(logfile=service_dir / "endpoint_storage.log")
self.conf_path = service_dir / "endpoint_storage.json"
service_dir = env.repo_dir / "object_storage"
super().__init__(logfile=service_dir / "object_storage.log")
self.conf_path = service_dir / "object_storage.json"
self.env = env
def base_url(self):
return json.loads(self.conf_path.read_text())["listen"]
def start(self, timeout_in_seconds: int | None = None):
self.env.neon_cli.endpoint_storage_start(timeout_in_seconds)
self.env.neon_cli.object_storage_start(timeout_in_seconds)
def stop(self, immediate: bool = False):
self.env.neon_cli.endpoint_storage_stop(immediate)
self.env.neon_cli.object_storage_stop(immediate)
def assert_no_errors(self):
assert_no_errors(self.logfile, "endpoint_storage", [])
assert_no_errors(self.logfile, "object_storage", [])
class NeonProxiedStorageController(NeonStorageController):

View File

@@ -65,7 +65,7 @@ def test_ro_replica_lag(
project = neon_api.create_project(pg_version)
project_id = project["project"]["id"]
log.info("Project ID: %s", project_id)
log.info("Primary endpoint ID: %s", project["endpoints"][0]["id"])
log.info("Primary endpoint ID: %s", project["project"]["endpoints"][0]["id"])
neon_api.wait_for_operation_to_finish(project_id)
error_occurred = False
try:
@@ -198,7 +198,7 @@ def test_replication_start_stop(
project = neon_api.create_project(pg_version)
project_id = project["project"]["id"]
log.info("Project ID: %s", project_id)
log.info("Primary endpoint ID: %s", project["endpoints"][0]["id"])
log.info("Primary endpoint ID: %s", project["project"]["endpoints"][0]["id"])
neon_api.wait_for_operation_to_finish(project_id)
try:
branch_id = project["branch"]["id"]

View File

@@ -1,7 +1,6 @@
from __future__ import annotations
import os
import platform
import shutil
import tarfile
from typing import TYPE_CHECKING
@@ -59,18 +58,7 @@ def test_remote_extensions(
extensions_endpoint = f"http://{host}:{port}/pg-ext-s3-gateway"
build_tag = os.environ.get("BUILD_TAG", "latest")
# We have decided to use the Go naming convention due to Kubernetes.
arch = platform.machine()
match arch:
case "aarch64":
arch = "arm64"
case "x86_64":
arch = "amd64"
case _:
pass
archive_route = f"{build_tag}/{arch}/v{pg_version}/extensions/test_extension.tar.zst"
archive_route = f"{build_tag}/v{pg_version}/extensions/test_extension.tar.zst"
tarball = test_output_dir / "test_extension.tar"
extension_dir = (
base_dir / "test_runner" / "regress" / "data" / "test_remote_extensions" / "test_extension"

View File

@@ -138,7 +138,7 @@ def test_cli_start_stop(neon_env_builder: NeonEnvBuilder):
env.neon_cli.pageserver_stop(env.pageserver.id)
env.neon_cli.safekeeper_stop()
env.neon_cli.storage_controller_stop(False)
env.neon_cli.endpoint_storage_stop(False)
env.neon_cli.object_storage_stop(False)
env.neon_cli.storage_broker_stop()
# Keep NeonEnv state up to date, it usually owns starting/stopping services
@@ -185,7 +185,7 @@ def test_cli_start_stop_multi(neon_env_builder: NeonEnvBuilder):
env.neon_cli.safekeeper_stop(neon_env_builder.safekeepers_id_start + 1)
env.neon_cli.safekeeper_stop(neon_env_builder.safekeepers_id_start + 2)
env.neon_cli.endpoint_storage_stop(False)
env.neon_cli.object_storage_stop(False)
# Stop this to get out of the way of the following `start`
env.neon_cli.storage_controller_stop(False)

View File

@@ -8,7 +8,7 @@ from jwcrypto import jwk, jwt
@pytest.mark.asyncio
async def test_endpoint_storage_insert_retrieve_delete(neon_simple_env: NeonEnv):
async def test_object_storage_insert_retrieve_delete(neon_simple_env: NeonEnv):
"""
Inserts, retrieves, and deletes test file using a JWT token
"""
@@ -31,7 +31,7 @@ async def test_endpoint_storage_insert_retrieve_delete(neon_simple_env: NeonEnv)
token.make_signed_token(key)
token = token.serialize()
base_url = env.endpoint_storage.base_url()
base_url = env.object_storage.base_url()
key = f"http://{base_url}/{tenant_id}/{timeline_id}/{endpoint_id}/key"
headers = {"Authorization": f"Bearer {token}"}
log.info(f"cache key url {key}")

View File

@@ -95,7 +95,7 @@ def test_storage_controller_smoke(
env.pageservers[1].start()
for sk in env.safekeepers:
sk.start()
env.endpoint_storage.start()
env.object_storage.start()
# The pageservers we started should have registered with the sharding service on startup
nodes = env.storage_controller.node_list()
@@ -347,7 +347,7 @@ def prepare_onboarding_env(
env = neon_env_builder.init_configs()
env.broker.start()
env.storage_controller.start()
env.endpoint_storage.start()
env.object_storage.start()
# This is the pageserver where we'll initially create the tenant. Run it in emergency
# mode so that it doesn't talk to storage controller, and do not register it.
@@ -1612,18 +1612,16 @@ def test_storage_controller_heartbeats(
env = neon_env_builder.init_configs()
env.start()
env.storage_controller.allowed_errors.extend(
[
# Default log allow list permits connection errors, but this test will use error responses on
# the utilization endpoint.
".*Call to node.*management API.*failed.*failpoint.*",
# The server starts listening to the socket before sending re-attach request,
# but it starts serving HTTP only when re-attach is completed.
# If re-attach is slow (last scenario), storcon's heartbeat requests will time out.
".*Call to node.*management API.*failed.* Timeout.*",
# We will intentionally cause reconcile errors
".*Reconcile error.*",
]
# Default log allow list permits connection errors, but this test will use error responses on
# the utilization endpoint.
env.storage_controller.allowed_errors.append(
".*Call to node.*management API.*failed.*failpoint.*"
)
# The server starts listening to the socket before sending re-attach request,
# but it starts serving HTTP only when re-attach is completed.
# If re-attach is slow (last scenario), storcon's heartbeat requests will time out.
env.storage_controller.allowed_errors.append(
".*Call to node.*management API.*failed.* Timeout.*"
)
# Initially we have two online pageservers
@@ -4242,63 +4240,6 @@ def test_storcon_create_delete_sk_down(
wait_until(timeline_deleted_on_sk)
@run_only_on_default_postgres("PG version is not interesting here")
@pytest.mark.parametrize("num_safekeepers", [1, 2, 3])
@pytest.mark.parametrize("deletetion_subject", [DeletionSubject.TENANT, DeletionSubject.TIMELINE])
def test_storcon_few_sk(
neon_env_builder: NeonEnvBuilder,
num_safekeepers: int,
deletetion_subject: DeletionSubject,
):
"""
Test that the storcon can create and delete tenants and timelines with a limited/special number of safekeepers
- num_safekeepers: number of safekeepers.
- deletion_subject: test that both single timeline and whole tenant deletion work.
"""
neon_env_builder.num_safekeepers = num_safekeepers
safekeeper_list = list(range(1, num_safekeepers + 1))
neon_env_builder.storage_controller_config = {
"timelines_onto_safekeepers": True,
}
env = neon_env_builder.init_start()
tenant_id = TenantId.generate()
timeline_id = TimelineId.generate()
env.create_tenant(tenant_id, timeline_id)
child_timeline_id = env.create_branch("child_of_main", tenant_id)
env.safekeepers[0].assert_log_contains(f"creating new timeline {tenant_id}/{timeline_id}")
config_lines = [
"neon.safekeeper_proto_version = 3",
]
with env.endpoints.create("main", tenant_id=tenant_id, config_lines=config_lines) as ep:
# endpoint should start.
ep.start(safekeeper_generation=1, safekeepers=safekeeper_list)
ep.safe_psql("CREATE TABLE IF NOT EXISTS t(key int, value text)")
with env.endpoints.create(
"child_of_main", tenant_id=tenant_id, config_lines=config_lines
) as ep:
# endpoint should start.
ep.start(safekeeper_generation=1, safekeepers=safekeeper_list)
ep.safe_psql("CREATE TABLE IF NOT EXISTS t(key int, value text)")
if deletetion_subject is DeletionSubject.TENANT:
env.storage_controller.pageserver_api().tenant_delete(tenant_id)
else:
env.storage_controller.pageserver_api().timeline_delete(tenant_id, child_timeline_id)
# ensure that there is log msgs for the third safekeeper too
def timeline_deleted_on_sk():
env.safekeepers[0].assert_log_contains(
f"deleting timeline {tenant_id}/{child_timeline_id} from disk"
)
wait_until(timeline_deleted_on_sk)
@pytest.mark.parametrize("wrong_az", [True, False])
def test_storage_controller_graceful_migration(neon_env_builder: NeonEnvBuilder, wrong_az: bool):
"""