rust/neon
1
0
Fork 0
mirror of https://github.com/neondatabase/neon.git synced 2026-06-01 04:20:39 +00:00
Code Issues Packages Projects Releases Wiki Activity

pageserver: fix rough edges of pageserver tracing (#11842)

Browse Source 
## Problem

There's a few rough edges around PS tracing.

## Summary of changes

* include compute request id in pageserver trace
* use the get page specific context for GET_REL_SIZE and GET_BATCH
* fix assertion in download layer trace


![image](https://github.com/user-attachments/assets/2ff6779c-7c2d-4102-8013-ada8203aa42f)
This commit is contained in:
Vlad Lazar
2025-05-07 11:13:26 +01:00
committed by GitHub
parent 608afc3055
commit 3cf5e1386c
3 changed files with 55 additions and 50 deletions
Download Patch File Download Diff File Expand all files Collapse all files

6
pageserver/src/page_service.rs
View File

@@ -1038,21 +1038,23 @@ impl PageServerHandler {
tracing::info_span!(
parent: &parent_span,
"handle_get_page_request",
request_id = %req.hdr.reqid,
rel = %req.rel,
blkno = %req.blkno,
req_lsn = %req.hdr.request_lsn,
not_modified_since_lsn = %req.hdr.not_modified_since
not_modified_since_lsn = %req.hdr.not_modified_since,
)
}};
($shard_id:expr) => {{
tracing::info_span!(
parent: &parent_span,
"handle_get_page_request",
request_id = %req.hdr.reqid,
rel = %req.rel,
blkno = %req.blkno,
req_lsn = %req.hdr.request_lsn,
not_modified_since_lsn = %req.hdr.not_modified_since,
shard_id = %$shard_id
shard_id = %$shard_id,
)
}};
}

64
pageserver/src/pgdatadir_mapping.rs
View File

@@ -40,7 +40,7 @@ use wal_decoder::serialized_batch::{SerializedValueBatch, ValueMeta};
use super::tenant::{PageReconstructError, Timeline};
use crate::aux_file;
use crate::context::{PerfInstrumentFutureExt, RequestContext};
use crate::context::{PerfInstrumentFutureExt, RequestContext, RequestContextBuilder};
use crate::keyspace::{KeySpace, KeySpaceAccum};
use crate::metrics::{
RELSIZE_CACHE_ENTRIES, RELSIZE_CACHE_HITS, RELSIZE_CACHE_MISSES, RELSIZE_CACHE_MISSES_OLD,
@@ -275,24 +275,30 @@ impl Timeline {
continue;
}
let nblocks = match self
.get_rel_size(*tag, Version::Lsn(lsn), &ctx)
.maybe_perf_instrument(&ctx, |crnt_perf_span| {
info_span!(
target: PERF_TRACE_TARGET,
parent: crnt_perf_span,
"GET_REL_SIZE",
reltag=%tag,
lsn=%lsn,
)
})
.await
{
Ok(nblocks) => nblocks,
Err(err) => {
result_slots[response_slot_idx].write(Err(err));
slots_filled += 1;
continue;
let nblocks = {
let ctx = RequestContextBuilder::from(&ctx)
.perf_span(|crnt_perf_span| {
info_span!(
target: PERF_TRACE_TARGET,
parent: crnt_perf_span,
"GET_REL_SIZE",
reltag=%tag,
lsn=%lsn,
)
})
.attached_child();
match self
.get_rel_size(*tag, Version::Lsn(lsn), &ctx)
.maybe_perf_instrument(&ctx, |crnt_perf_span| crnt_perf_span.clone())
.await
{
Ok(nblocks) => nblocks,
Err(err) => {
result_slots[response_slot_idx].write(Err(err));
slots_filled += 1;
continue;
}
}
};
@@ -308,6 +314,17 @@ impl Timeline {
let key = rel_block_to_key(*tag, *blknum);
let ctx = RequestContextBuilder::from(&ctx)
.perf_span(|crnt_perf_span| {
info_span!(
target: PERF_TRACE_TARGET,
parent: crnt_perf_span,
"GET_BATCH",
batch_size = %page_count,
)
})
.attached_child();
let key_slots = keys_slots.entry(key).or_default();
key_slots.push((response_slot_idx, ctx));
@@ -323,14 +340,7 @@ impl Timeline {
let query = VersionedKeySpaceQuery::scattered(query);
let res = self
.get_vectored(query, io_concurrency, ctx)
.maybe_perf_instrument(ctx, |current_perf_span| {
info_span!(
target: PERF_TRACE_TARGET,
parent: current_perf_span,
"GET_BATCH",
batch_size = %page_count,
)
})
.maybe_perf_instrument(ctx, |current_perf_span| current_perf_span.clone())
.await;
match res {

35
pageserver/src/tenant/storage_layer/layer.rs
View File

@@ -23,7 +23,7 @@ use super::{
LayerVisibilityHint, PerfInstrumentFutureExt, PersistentLayerDesc, ValuesReconstructState,
};
use crate::config::PageServerConf;
use crate::context::{DownloadBehavior, RequestContext, RequestContextBuilder};
use crate::context::{RequestContext, RequestContextBuilder};
use crate::span::debug_assert_current_span_has_tenant_and_timeline_id;
use crate::task_mgr::TaskKind;
use crate::tenant::Timeline;
@@ -1076,24 +1076,17 @@ impl LayerInner {
return Err(DownloadError::DownloadRequired);
}
let ctx = if ctx.has_perf_span() {
let dl_ctx = RequestContextBuilder::from(ctx)
.task_kind(TaskKind::LayerDownload)
.download_behavior(DownloadBehavior::Download)
.root_perf_span(|| {
info_span!(
target: PERF_TRACE_TARGET,
"DOWNLOAD_LAYER",
layer = %self,
reason = %reason
)
})
.detached_child();
ctx.perf_follows_from(&dl_ctx);
dl_ctx
} else {
ctx.attached_child()
};
let ctx = RequestContextBuilder::from(ctx)
.perf_span(|crnt_perf_span| {
info_span!(
target: PERF_TRACE_TARGET,
parent: crnt_perf_span,
"DOWNLOAD_LAYER",
layer = %self,
reason = %reason,
)
})
.attached_child();
async move {
tracing::info!(%reason, "downloading on-demand");
@@ -1101,7 +1094,7 @@ impl LayerInner {
let init_cancelled = scopeguard::guard((), |_| LAYER_IMPL_METRICS.inc_init_cancelled());
let res = self
.download_init_and_wait(timeline, permit, ctx.attached_child())
.maybe_perf_instrument(&ctx, |crnt_perf_span| crnt_perf_span.clone())
.maybe_perf_instrument(&ctx, |current_perf_span| current_perf_span.clone())
.await?;
scopeguard::ScopeGuard::into_inner(init_cancelled);
@@ -1709,7 +1702,7 @@ impl DownloadError {
}
}
#[derive(Debug, PartialEq)]
#[derive(Debug, PartialEq, Copy, Clone)]
pub(crate) enum NeedsDownload {
NotFound,
NotFile(std::fs::FileType),