storage scrubber: GC ancestor shard layers (#8196)

## Problem

After a shard split, the pageserver leaves the ancestor shard's content
in place. It may be referenced by child shards, but eventually child
shards will de-reference most ancestor layers as they write their own
data and do GC. We would like to eventually clean up those ancestor
layers to reclaim space.

## Summary of changes

- Extend the physical GC command with `--mode=full`, which includes
cleaning up unreferenced ancestor shard layers
- Add test `test_scrubber_physical_gc_ancestors`
- Remove colored log output: in testing this is irritating ANSI code
spam in logs, and in interactive use doesn't add much.
- Refactor storage controller API client code out of storcon_client into
a `storage_controller/client` crate
- During physical GC of ancestors, call into the storage controller to
check that the latest shards seen in S3 reflect the latest state of the
tenant, and there is no shard split in progress.
This commit is contained in:
John Spray
2024-07-19 17:07:59 +01:00
committed by GitHub
parent 16071e57c6
commit 44781518d0
24 changed files with 905 additions and 191 deletions

View File

@@ -17,6 +17,7 @@ pageserver_client.workspace = true
reqwest.workspace = true
serde.workspace = true
serde_json = { workspace = true, features = ["raw_value"] }
storage_controller_client.workspace = true
thiserror.workspace = true
tokio.workspace = true
tracing.workspace = true

View File

@@ -14,15 +14,15 @@ use pageserver_api::{
},
shard::{ShardStripeSize, TenantShardId},
};
use pageserver_client::mgmt_api::{self, ResponseErrorMessageExt};
use pageserver_client::mgmt_api::{self};
use reqwest::{Method, StatusCode, Url};
use serde::{de::DeserializeOwned, Serialize};
use utils::id::{NodeId, TenantId};
use pageserver_api::controller_api::{
NodeConfigureRequest, NodeRegisterRequest, NodeSchedulingPolicy, PlacementPolicy,
TenantShardMigrateRequest, TenantShardMigrateResponse,
};
use storage_controller_client::control_api::Client;
#[derive(Subcommand, Debug)]
enum Command {
@@ -249,64 +249,6 @@ impl FromStr for NodeAvailabilityArg {
}
}
struct Client {
base_url: Url,
jwt_token: Option<String>,
client: reqwest::Client,
}
impl Client {
fn new(base_url: Url, jwt_token: Option<String>) -> Self {
Self {
base_url,
jwt_token,
client: reqwest::ClientBuilder::new()
.build()
.expect("Failed to construct http client"),
}
}
/// Simple HTTP request wrapper for calling into storage controller
async fn dispatch<RQ, RS>(
&self,
method: Method,
path: String,
body: Option<RQ>,
) -> mgmt_api::Result<RS>
where
RQ: Serialize + Sized,
RS: DeserializeOwned + Sized,
{
// The configured URL has the /upcall path prefix for pageservers to use: we will strip that out
// for general purpose API access.
let url = Url::from_str(&format!(
"http://{}:{}/{path}",
self.base_url.host_str().unwrap(),
self.base_url.port().unwrap()
))
.unwrap();
let mut builder = self.client.request(method, url);
if let Some(body) = body {
builder = builder.json(&body)
}
if let Some(jwt_token) = &self.jwt_token {
builder = builder.header(
reqwest::header::AUTHORIZATION,
format!("Bearer {jwt_token}"),
);
}
let response = builder.send().await.map_err(mgmt_api::Error::ReceiveBody)?;
let response = response.error_from_body().await?;
response
.json()
.await
.map_err(pageserver_client::mgmt_api::Error::ReceiveBody)
}
}
#[tokio::main]
async fn main() -> anyhow::Result<()> {
let cli = Cli::parse();