Compare commits

..

1 Commits

Author SHA1 Message Date
Stas Kelvich
892723cc7e Local auth renew proxy 2025-03-22 03:31:54 +02:00
20 changed files with 200 additions and 338 deletions

View File

@@ -41,14 +41,6 @@ storage:
ansible_host: i-051642d372c0a4f32
pageserver-3.us-west-2.aws.neon.tech:
ansible_host: i-00c3844beb9ad1c6b
pageserver-4.us-west-2.aws.neon.tech:
ansible_host: i-013263dd1c239adcc
pageserver-5.us-west-2.aws.neon.tech:
ansible_host: i-00ca6417c7bf96820
pageserver-6.us-west-2.aws.neon.tech:
ansible_host: i-01cdf7d2bc1433b6a
pageserver-7.us-west-2.aws.neon.tech:
ansible_host: i-02eec9b40617db5bc
safekeepers:
hosts:
@@ -58,15 +50,4 @@ storage:
ansible_host: i-074682f9d3c712e7c
safekeeper-2.us-west-2.aws.neon.tech:
ansible_host: i-042b7efb1729d7966
safekeeper-3.us-west-2.aws.neon.tech:
ansible_host: i-089f6b9ef426dff76
safekeeper-4.us-west-2.aws.neon.tech:
ansible_host: i-0fe6bf912c4710c82
safekeeper-5.us-west-2.aws.neon.tech:
ansible_host: i-0a83c1c46d2b4e409
safekeeper-6.us-west-2.aws.neon.tech:
ansible_host: i-0fef5317b8fdc9f8d
safekeeper-7.us-west-2.aws.neon.tech:
ansible_host: i-0be739190d4289bf9
safekeeper-8.us-west-2.aws.neon.tech:
ansible_host: i-00e851803669e5cfe

View File

@@ -35,8 +35,6 @@ storage:
hosts:
pageserver-0.eu-west-1.aws.neon.build:
ansible_host: i-01d496c5041c7f34c
pageserver-1.eu-west-1.aws.neon.build:
ansible_host: i-0e8013e239ce3928c
safekeepers:
hosts:
@@ -46,15 +44,3 @@ storage:
ansible_host: i-06969ee1bf2958bfc
safekeeper-2.eu-west-1.aws.neon.build:
ansible_host: i-087892e9625984a0b
safekeeper-3.eu-west-1.aws.neon.build:
ansible_host: i-0a6f91660e99e8891
safekeeper-4.eu-west-1.aws.neon.build:
ansible_host: i-0012e309e28e7c249
safekeeper-5.eu-west-1.aws.neon.build:
ansible_host: i-085a2b1193287b32e
safekeeper-6.eu-west-1.aws.neon.build:
ansible_host: i-0c713248465ed0fbd
safekeeper-7.eu-west-1.aws.neon.build:
ansible_host: i-02ad231aed2a80b7a
safekeeper-8.eu-west-1.aws.neon.build:
ansible_host: i-0dbbd8ffef66efda8

View File

@@ -1,19 +0,0 @@
useCertManager: true
replicaCount: 3
exposedService:
# exposedService.port -- Exposed Service proxy port
port: 4432
annotations:
external-dns.alpha.kubernetes.io/hostname: "*.snirouter.alpha.eu-central-1.internal.aws.neon.build"
settings:
domain: "*.snirouter.alpha.eu-central-1.internal.aws.neon.build"
sentryEnvironment: "staging"
imagePullSecrets:
- name: docker-hub-neon
metrics:
enabled: false

View File

@@ -1,19 +0,0 @@
useCertManager: true
replicaCount: 3
exposedService:
# exposedService.port -- Exposed Service proxy port
port: 4432
annotations:
external-dns.alpha.kubernetes.io/hostname: "*.snirouter.zeta.eu-west-1.internal.aws.neon.build"
settings:
domain: "*.snirouter.zeta.eu-west-1.internal.aws.neon.build"
sentryEnvironment: "staging"
imagePullSecrets:
- name: docker-hub-neon
metrics:
enabled: false

View File

@@ -1,19 +0,0 @@
useCertManager: true
replicaCount: 3
exposedService:
# exposedService.port -- Exposed Service proxy port
port: 4432
annotations:
external-dns.alpha.kubernetes.io/hostname: "*.snirouter.beta.us-east-2.internal.aws.neon.build"
settings:
domain: "*.snirouter.beta.us-east-2.internal.aws.neon.build"
sentryEnvironment: "staging"
imagePullSecrets:
- name: docker-hub-neon
metrics:
enabled: false

View File

@@ -1,19 +0,0 @@
useCertManager: true
replicaCount: 3
exposedService:
# exposedService.port -- Exposed Service proxy port
port: 4432
annotations:
external-dns.alpha.kubernetes.io/hostname: "*.snirouter.epsilon.ap-southeast-1.internal.aws.neon.tech"
settings:
domain: "*.snirouter.epsilon.ap-southeast-1.internal.aws.neon.tech"
sentryEnvironment: "production"
imagePullSecrets:
- name: docker-hub-neon
metrics:
enabled: false

View File

@@ -1,19 +0,0 @@
useCertManager: true
replicaCount: 3
exposedService:
# exposedService.port -- Exposed Service proxy port
port: 4432
annotations:
external-dns.alpha.kubernetes.io/hostname: "*.snirouter.gamma.eu-central-1.internal.aws.neon.tech"
settings:
domain: "*.snirouter.gamma.eu-central-1.internal.aws.neon.tech"
sentryEnvironment: "production"
imagePullSecrets:
- name: docker-hub-neon
metrics:
enabled: false

View File

@@ -1,19 +0,0 @@
useCertManager: true
replicaCount: 3
exposedService:
# exposedService.port -- Exposed Service proxy port
port: 4432
annotations:
external-dns.alpha.kubernetes.io/hostname: "*.snirouter.theta.us-east-1.internal.aws.neon.tech"
settings:
domain: "*.snirouter.theta.us-east-1.internal.aws.neon.tech"
sentryEnvironment: "production"
imagePullSecrets:
- name: docker-hub-neon
metrics:
enabled: false

View File

@@ -1,19 +0,0 @@
useCertManager: true
replicaCount: 3
exposedService:
# exposedService.port -- Exposed Service proxy port
port: 4432
annotations:
external-dns.alpha.kubernetes.io/hostname: "*.snirouter.delta.us-east-2.internal.aws.neon.tech"
settings:
domain: "*.snirouter.delta.us-east-2.internal.aws.neon.tech"
sentryEnvironment: "production"
imagePullSecrets:
- name: docker-hub-neon
metrics:
enabled: false

View File

@@ -1,19 +0,0 @@
useCertManager: true
replicaCount: 3
exposedService:
# exposedService.port -- Exposed Service proxy port
port: 4432
annotations:
external-dns.alpha.kubernetes.io/hostname: "*.snirouter.eta.us-west-2.internal.aws.neon.tech"
settings:
domain: "*.snirouter.eta.us-west-2.internal.aws.neon.tech"
sentryEnvironment: "production"
imagePullSecrets:
- name: docker-hub-neon
metrics:
enabled: false

View File

@@ -27,11 +27,6 @@ on:
required: true
type: boolean
default: true
deployPgSniRouter:
description: 'Deploy pg-sni-router'
required: true
type: boolean
default: true
env:
AWS_ACCESS_KEY_ID: ${{ secrets.AWS_ACCESS_KEY_DEV }}
@@ -232,49 +227,3 @@ jobs:
- name: Cleanup helm folder
run: rm -rf ~/.cache
deploy-pg-sni-router:
runs-on: [ self-hosted, gen3, small ]
container: 369495373322.dkr.ecr.eu-central-1.amazonaws.com/ansible:pinned
if: inputs.deployPgSniRouter
defaults:
run:
shell: bash
strategy:
matrix:
include:
- target_region: us-east-2
target_cluster: dev-us-east-2-beta
- target_region: eu-west-1
target_cluster: dev-eu-west-1-zeta
- target_region: eu-central-1
target_cluster: dev-eu-central-1-alpha
environment:
name: dev-${{ matrix.target_region }}
steps:
- name: Checkout
uses: actions/checkout@v3
with:
submodules: true
fetch-depth: 0
ref: ${{ inputs.branch }}
- name: Configure AWS Credentials
uses: aws-actions/configure-aws-credentials@v1-node16
with:
role-to-assume: arn:aws:iam::369495373322:role/github-runner
aws-region: eu-central-1
role-skip-session-tagging: true
role-duration-seconds: 1800
- name: Configure environment
run: |
helm repo add neondatabase https://neondatabase.github.io/helm-charts
aws --region ${{ matrix.target_region }} eks update-kubeconfig --name ${{ matrix.target_cluster }}
- name: Deploy pg-sni-router
run:
helm upgrade neon-pg-sni-router neondatabase/neon-pg-sni-router --namespace neon-pg-sni-router --create-namespace --install --atomic -f .github/helm-values/${{ matrix.target_cluster }}.pg-sni-router.yaml --set image.tag=${{ inputs.dockerTag }} --set settings.sentryUrl=${{ secrets.SENTRY_URL_BROKER }} --wait --timeout 15m0s
- name: Cleanup helm folder
run: rm -rf ~/.cache

View File

@@ -27,11 +27,6 @@ on:
required: true
type: boolean
default: true
deployPgSniRouter:
description: 'Deploy pg-sni-router'
required: true
type: boolean
default: true
disclamerAcknowledged:
description: 'I confirm that there is an emergency and I can not use regular release workflow'
required: true
@@ -176,42 +171,3 @@ jobs:
- name: Deploy storage-broker
run:
helm upgrade neon-storage-broker-lb neondatabase/neon-storage-broker --namespace neon-storage-broker-lb --create-namespace --install --atomic -f .github/helm-values/${{ matrix.target_cluster }}.neon-storage-broker.yaml --set image.tag=${{ inputs.dockerTag }} --set settings.sentryUrl=${{ secrets.SENTRY_URL_BROKER }} --wait --timeout 5m0s
deploy-pg-sni-router:
runs-on: prod
container: 093970136003.dkr.ecr.eu-central-1.amazonaws.com/ansible:latest
if: inputs.deployPgSniRouter && inputs.disclamerAcknowledged
defaults:
run:
shell: bash
strategy:
matrix:
include:
- target_region: us-east-2
target_cluster: prod-us-east-2-delta
- target_region: us-west-2
target_cluster: prod-us-west-2-eta
- target_region: eu-central-1
target_cluster: prod-eu-central-1-gamma
- target_region: ap-southeast-1
target_cluster: prod-ap-southeast-1-epsilon
- target_region: us-east-1
target_cluster: prod-us-east-1-theta
environment:
name: prod-${{ matrix.target_region }}
steps:
- name: Checkout
uses: actions/checkout@v3
with:
submodules: true
fetch-depth: 0
ref: ${{ inputs.branch }}
- name: Configure environment
run: |
helm repo add neondatabase https://neondatabase.github.io/helm-charts
aws --region ${{ matrix.target_region }} eks update-kubeconfig --name ${{ matrix.target_cluster }}
- name: Deploy pg-sni-router
run:
helm upgrade neon-pg-sni-router neondatabase/neon-pg-sni-router --namespace neon-pg-sni-router --create-namespace --install --atomic -f .github/helm-values/${{ matrix.target_cluster }}.pg-sni-router.yaml --set image.tag=${{ inputs.dockerTag }} --set settings.sentryUrl=${{ secrets.SENTRY_URL_BROKER }} --wait --timeout 15m0s

3
Cargo.lock generated
View File

@@ -3960,7 +3960,8 @@ dependencies = [
[[package]]
name = "sharded-slab"
version = "0.1.4"
source = "git+https://github.com/neondatabase/sharded-slab.git?rev=98d16753ab01c61f0a028de44167307a00efea00#98d16753ab01c61f0a028de44167307a00efea00"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "900fba806f70c630b0a382d0d825e17a0f19fcd059a2ade1ff237bcddf446b31"
dependencies = [
"lazy_static",
]

View File

@@ -161,17 +161,11 @@ rstest = "0.17"
tempfile = "3.4"
tonic-build = "0.9"
[patch.crates-io]
# This is only needed for proxy's tests.
# TODO: we should probably fork `tokio-postgres-rustls` instead.
[patch.crates-io]
tokio-postgres = { git = "https://github.com/neondatabase/rust-postgres.git", rev="0bc41d8503c092b040142214aac3cf7d11d0c19f" }
# Changes the MAX_THREADS limit from 4096 to 32768.
# This is a temporary workaround for using tracing from many threads in safekeepers code,
# until async safekeepers patch is merged to the main.
sharded-slab = { git = "https://github.com/neondatabase/sharded-slab.git", rev="98d16753ab01c61f0a028de44167307a00efea00" }
################# Binary contents sections
[profile.release]

View File

@@ -44,15 +44,7 @@ COPY --chown=nonroot . .
# Show build caching stats to check if it was used in the end.
# Has to be the part of the same RUN since cachepot daemon is killed in the end of this RUN, losing the compilation stats.
RUN set -e \
&& mold -run cargo build \
--bin pg_sni_router \
--bin pageserver \
--bin pageserver_binutils \
--bin draw_timeline_dir \
--bin safekeeper \
--bin storage_broker \
--bin proxy \
--locked --release \
&& mold -run cargo build --bin pageserver --bin pageserver_binutils --bin draw_timeline_dir --bin safekeeper --bin storage_broker --bin proxy --locked --release \
&& cachepot -s
# Build final image
@@ -71,7 +63,6 @@ RUN set -e \
&& useradd -d /data neon \
&& chown -R neon:neon /data
COPY --from=build --chown=neon:neon /home/nonroot/target/release/pg_sni_router /usr/local/bin
COPY --from=build --chown=neon:neon /home/nonroot/target/release/pageserver /usr/local/bin
COPY --from=build --chown=neon:neon /home/nonroot/target/release/pageserver_binutils /usr/local/bin
COPY --from=build --chown=neon:neon /home/nonroot/target/release/draw_timeline_dir /usr/local/bin

View File

@@ -287,33 +287,14 @@ impl EvictionsWithLowResidenceDuration {
let Some(_counter) = self.counter.take() else {
return;
};
let threshold = Self::threshold_label_value(self.threshold);
let removed = EVICTIONS_WITH_LOW_RESIDENCE_DURATION.remove_label_values(&[
tenant_id,
timeline_id,
self.data_source,
&threshold,
]);
match removed {
Err(e) => {
// this has been hit in staging as
// <https://neondatabase.sentry.io/issues/4142396994/>, but we don't know how.
// because we can be in the drop path already, don't risk:
// - "double-panic => illegal instruction" or
// - future "drop panick => abort"
//
// so just nag: (the error has the labels)
tracing::warn!("failed to remove EvictionsWithLowResidenceDuration, it was already removed? {e:#?}");
}
Ok(()) => {
// to help identify cases where we double-remove the same values, let's log all
// deletions?
tracing::info!("removed EvictionsWithLowResidenceDuration with {tenant_id}, {timeline_id}, {}, {threshold}", self.data_source);
}
}
EVICTIONS_WITH_LOW_RESIDENCE_DURATION
.remove_label_values(&[
tenant_id,
timeline_id,
self.data_source,
&Self::threshold_label_value(self.threshold),
])
.expect("we own the metric, no-one else should remove it");
}
}

View File

@@ -352,7 +352,7 @@ impl PageServerHandler {
tenant_id: TenantId,
timeline_id: TimelineId,
ctx: RequestContext,
) -> Result<(), QueryError>
) -> anyhow::Result<()>
where
IO: AsyncRead + AsyncWrite + Send + Sync + Unpin,
{
@@ -398,9 +398,7 @@ impl PageServerHandler {
Some(FeMessage::CopyData(bytes)) => bytes,
Some(FeMessage::Terminate) => break,
Some(m) => {
return Err(QueryError::Other(anyhow::anyhow!(
"unexpected message: {m:?} during COPY"
)));
anyhow::bail!("unexpected message: {m:?} during COPY");
}
None => break, // client disconnected
};

View File

@@ -0,0 +1,183 @@
use std::{net::SocketAddr, sync::Arc};
use tokio::{io::AsyncWriteExt, net::TcpListener};
use anyhow::Context;
use clap::{self, Arg};
use futures::TryFutureExt;
use proxy::{
auth::{self, AuthFlow},
cancellation::CancelMap,
compute::ConnCfg,
console::messages::MetricsAuxInfo,
};
use tokio::io::{AsyncRead, AsyncWrite};
use tokio_postgres::config::SslMode;
use tokio_util::sync::CancellationToken;
use utils::project_git_version;
use tracing::{error, info, warn};
project_git_version!(GIT_VERSION);
fn cli() -> clap::Command {
clap::Command::new("Auth renew proxy")
.disable_help_flag(true)
.version(GIT_VERSION)
.arg(
Arg::new("listen")
.short('l')
.long("listen")
.help("listen for incoming client connections on ip:port")
.default_value("127.0.0.1:4432"),
)
.arg(
Arg::new("dest-host")
.long("dest-host")
.help("destination hosts")
.required(true),
)
.arg(
Arg::new("dest-port")
.long("dest-port")
.help("destination port")
.default_value("5432"),
)
}
#[tokio::main]
async fn main() -> anyhow::Result<()> {
let _logging_guard = proxy::logging::init().await?;
let _panic_hook_guard = utils::logging::replace_panic_hook_with_tracing_panic_hook();
let args = cli().get_matches();
let dest_host: String = args.get_one::<String>("dest-host").unwrap().parse()?;
let dest_port: u16 = args.get_one::<String>("dest-port").unwrap().parse()?;
let listen_address: SocketAddr = args.get_one::<String>("listen").unwrap().parse()?;
// Start listening for incoming client connections
info!("Starting proxy on {listen_address}");
let proxy_listener = TcpListener::bind(listen_address).await?;
let cancellation_token = CancellationToken::new();
let main = proxy::flatten_err(tokio::spawn(task_main(
Arc::new(dest_host),
dest_port,
proxy_listener,
cancellation_token.clone(),
)));
let signals_task = proxy::flatten_err(tokio::spawn(proxy::handle_signals(cancellation_token)));
tokio::select! {
res = main => { res?; },
res = signals_task => { res?; },
}
Ok(())
}
async fn task_main(
dest_host: Arc<String>,
dest_port: u16,
listener: tokio::net::TcpListener,
cancellation_token: CancellationToken,
) -> anyhow::Result<()> {
scopeguard::defer! {
info!("proxy has shut down");
}
// When set for the server socket, the keepalive setting
// will be inherited by all accepted client sockets.
socket2::SockRef::from(&listener).set_keepalive(true)?;
let mut connections = tokio::task::JoinSet::new();
let cancel_map = Arc::new(CancelMap::default());
loop {
tokio::select! {
accept_result = listener.accept() => {
let (socket, peer_addr) = accept_result?;
info!("accepted postgres client connection from {peer_addr}");
let cancel_map = Arc::clone(&cancel_map);
let dest_host = Arc::clone(&dest_host);
connections.spawn(
async move {
info!("spawned a task for {peer_addr}");
socket
.set_nodelay(true)
.context("failed to set socket option")?;
handle_client(dest_host, dest_port, &cancel_map, socket).await
}
.unwrap_or_else(|e| {
// Acknowledge that the task has finished with an error.
error!("per-client task finished with an error: {e:#}");
}),
);
}
_ = cancellation_token.cancelled() => {
drop(listener);
break;
}
}
}
// Drain connections
while let Some(res) = connections.join_next().await {
if let Err(e) = res {
if !e.is_panic() && !e.is_cancelled() {
warn!("unexpected error from joined connection task: {e:?}");
}
}
}
Ok(())
}
async fn handle_client(
dest_host: Arc<String>,
dest_port: u16,
cancel_map: &CancelMap,
stream: impl AsyncRead + AsyncWrite + Unpin,
) -> anyhow::Result<()> {
let do_handshake = proxy::proxy::handshake(stream, None, cancel_map);
let (mut stream, params) = match do_handshake.await? {
Some(x) => x,
None => return Ok(()), // it's a cancellation request
};
// Here we force plain test auth for the client and using received password to authenticate
// to the destination server. Instead we can always trust the client and take the password / JWT
// each time we get a connection.
let password = AuthFlow::new(&mut stream)
.begin(auth::CleartextPassword)
.await?
.authenticate()
.await?;
let mut conn_cfg = ConnCfg::new();
conn_cfg.set_startup_params(&params);
conn_cfg.password(password);
conn_cfg.host(dest_host.as_str());
conn_cfg.port(dest_port);
conn_cfg.ssl_mode(SslMode::Require);
info!("destination: {:?}:{}", dest_host, dest_port);
let mut conn = conn_cfg
.connect(false)
.or_else(|e| stream.throw_error(e))
.await?;
cancel_map
.with_session(|session| async {
proxy::proxy::prepare_client_connection(&conn, false, session, &mut stream).await?;
let (stream, read_buf) = stream.into_inner();
conn.stream.write_all(&read_buf).await?;
let metrics_aux: MetricsAuxInfo = Default::default();
proxy::proxy::proxy_pass(stream, conn.stream, &metrics_aux).await
})
.await
}

View File

@@ -1,4 +1,3 @@
use tracing_opentelemetry::OpenTelemetryLayer;
use tracing_subscriber::{
filter::{EnvFilter, LevelFilter},
prelude::*,
@@ -22,13 +21,8 @@ pub async fn init() -> anyhow::Result<LoggingGuard> {
.with_writer(std::io::stderr)
.with_target(false);
let otlp_layer = tracing_utils::init_tracing("proxy")
.await
.map(OpenTelemetryLayer::new);
tracing_subscriber::registry()
.with(env_filter)
.with(otlp_layer)
.with(fmt_layer)
.try_init()?;

View File

@@ -213,7 +213,7 @@ async fn handle_client(
/// It's easier to work with owned `stream` here as we need to upgrade it to TLS;
/// we also take an extra care of propagating only the select handshake errors to client.
#[tracing::instrument(skip_all)]
async fn handshake<S: AsyncRead + AsyncWrite + Unpin>(
pub async fn handshake<S: AsyncRead + AsyncWrite + Unpin>(
stream: S,
mut tls: Option<&TlsConfig>,
cancel_map: &CancelMap,
@@ -350,7 +350,7 @@ async fn connect_to_compute(
/// Finish client connection initialization: confirm auth success, send params, etc.
#[tracing::instrument(skip_all)]
async fn prepare_client_connection(
pub async fn prepare_client_connection(
node: &compute::PostgresConnection,
reported_auth_ok: bool,
session: cancellation::Session<'_>,