disk-usage-based layer eviction

This patch adds a pageserver-global background loop that evicts layers
in response to a shortage of available bytes in the $repo/tenants
directory's filesystem.

The loop runs periodically at a configurable `period`.

Each loop iteration uses `statvfs` to determine filesystem-level space
usage.  It compares the returned usage data against two different types
of thresholds. The iteration tries to evict layers until app-internal
accounting says we should be below the thresholds.  We cross-check this
internal accounting with the real world by making another `statvfs` at
the end of the iteration.  We're good if that second statvfs shows that
we're _actually_ below the configured thresholds.  If we're still above
one or more thresholds, we emit a warning log message, leaving it to the
operator to investigate further.

There are two thresholds: `max_usage_pct` is the relative available
space, expressed in percent of the total filesystem space. If the actual
usage is higher, the threshold is exceeded.  `min_avail_bytes` is the
absolute available space in bytes. If the actual usage is lower, the
threshold is exceeded.

The iteration evicts layers in LRU fashion with a reservation of up to
`min_resident_size` bytes of the most recent layers per tenant.
The layers not part of the per-tenant reservation are evicted
least-recently-used first until we're below all thresholds.
If the above doesn't relieve enough pressure, we fall back to Global LRU.

In addition to the loop, there is also an HTTP endpoint to perform
one loop iteration synchronous to the request.
The endpoint takes an absolute number of bytes that the iteration
needs to evict before pressure is relieved.
The tests use this endpoint, which is a great simplification over
setting up loopback-mounts in the tests, which would be required to
test the statvfs part of the implementation.
We will rely on manual testing in staging to test the statvfs parts.

The HTTP endpoint is also handy in emergencies where an operator wants
the pageserver to evict a given amount of space _now.
Hence, it's arguments documented in openapi_spec.yml.
The response type isn't documented though because we don't consider
it stable. The endpoint should _not_ be used by Console.

Co-authored-by: Joonas Koivunen <joonas@neon.tech>

fixes https://github.com/neondatabase/neon/issues/3728
This commit is contained in:
Christian Schwarz
2023-03-20 16:46:23 +01:00
parent 881356c417
commit 5aef192bf2
20 changed files with 1373 additions and 14 deletions

View File

@@ -27,6 +27,31 @@ paths:
id:
type: integer
/v1/disk_usage_eviction/run:
put:
description: Do an iteration of disk-usage-based eviction to evict a given amount of disk space.
security: []
requestBody:
content:
application/json:
schema:
type: object
required:
- wanted_trimmed_bytes
properties:
wanted_trimmed_bytes:
type: integer
responses:
"200":
description: |
The run completed.
This does not necessarily mean that we actually evicted `wanted_trimmed_bytes`.
Examine the returned object for detail, or, just watch the actual effect of the call using `du` or `df`.
content:
application/json:
schema:
type: object
/v1/tenant/{tenant_id}:
parameters:
- name: tenant_id

View File

@@ -746,6 +746,8 @@ async fn tenant_create_handler(mut request: Request<Body>) -> Result<Response<Bo
);
}
tenant_conf.min_resident_size_override = request_data.min_resident_size_override;
let target_tenant_id = request_data
.new_tenant_id
.map(TenantId::from)
@@ -877,6 +879,8 @@ async fn update_tenant_config_handler(
);
}
tenant_conf.min_resident_size_override = request_data.min_resident_size_override;
let state = get_state(&request);
mgr::set_new_tenant_config(state.conf, tenant_conf, tenant_id)
.instrument(info_span!("tenant_config", tenant = ?tenant_id))
@@ -1171,6 +1175,105 @@ pub fn make_router(
"/v1/tenant/:tenant_id/timeline/:timeline_id/layer/:layer_file_name",
|r| RequestSpan(evict_timeline_layer_handler).handle(r),
)
.put("/v1/disk_usage_eviction/run", disk_usage_eviction_run)
.put(
"/v1/tenant/:tenant_id/break",
testing_api!("set tenant state to broken", handle_tenant_break),
)
.get("/v1/panic", |r| RequestSpan(always_panic_handler).handle(r))
.any(handler_404))
}
/// Testing helper to transition a tenant to [`crate::tenant::TenantState::Broken`].
#[cfg(feature = "testing")]
async fn handle_tenant_break(r: Request<Body>) -> Result<Response<Body>, ApiError> {
use std::str::FromStr;
let tenant_id = get_request_param(&r, "tenant_id")?;
let tenant_id = TenantId::from_str(tenant_id).map_err(|e| ApiError::BadRequest(e.into()))?;
let tenant = crate::tenant::mgr::get_tenant(tenant_id, true)
.await
.map_err(|_| ApiError::Conflict(String::from("no active tenant found")))?;
tenant.set_broken("broken from test");
json_response(StatusCode::OK, ())
}
async fn disk_usage_eviction_run(mut r: Request<Body>) -> Result<Response<Body>, ApiError> {
check_permission(&r, None)?;
#[derive(serde::Deserialize)]
struct Config {
/// How much to trim at minimum
wanted_trimmed_bytes: u64,
}
#[derive(Debug, Clone, Copy, serde::Serialize)]
struct Usage {
wanted_trimmed_bytes: u64,
freed_bytes: u64,
}
impl crate::disk_usage_eviction_task::Usage for Usage {
fn has_pressure(&self) -> bool {
self.wanted_trimmed_bytes > self.freed_bytes
}
fn add_available_bytes(&mut self, bytes: u64) {
self.freed_bytes += bytes;
}
}
let config = json_request::<Config>(&mut r)
.await
.map_err(|_| ApiError::BadRequest(anyhow::anyhow!("invalid JSON body")))?;
let usage = Usage {
wanted_trimmed_bytes: config.wanted_trimmed_bytes,
freed_bytes: 0,
};
use crate::task_mgr::MGMT_REQUEST_RUNTIME;
let (tx, rx) = tokio::sync::oneshot::channel();
let state = get_state(&r);
let Some(storage) = state.remote_storage.clone() else {
return Err(ApiError::InternalServerError(anyhow::anyhow!(
"remote storage not configured, cannot run eviction iteration"
)))
};
let cancel = CancellationToken::new();
let child_cancel = cancel.clone();
let _g = cancel.drop_guard();
crate::task_mgr::spawn(
MGMT_REQUEST_RUNTIME.handle(),
TaskKind::DiskUsageEviction,
None,
None,
"ondemand disk usage eviction",
false,
async move {
let res = crate::disk_usage_eviction_task::disk_usage_eviction_task_iteration_impl(
&storage,
usage,
&child_cancel,
)
.await;
info!(?res, "disk_usage_eviction_task_iteration_impl finished");
let _ = tx.send(res);
Ok(())
}
.in_current_span(),
);
let response = rx.await.unwrap().map_err(ApiError::InternalServerError)?;
json_response(StatusCode::OK, response)
}