wave13: pyroscope-0001 UNDF-1301 (47x-438x GetBlockStats) + 5 clean-scan additions
Flagship: pyroscope PhlareDB.GetBlockStats slices.Contains per block (O(B*U) -> O(B+U)). Long-retention tenants with 5k-10k blocks pay 1M+ membership checks per block-stats RPC. Set hoist: 438x at B=10k U=1k. Wave 13 honor roll: lima, apollo-server, act, firecracker, nix. Cumulative: 59 projects.
This commit is contained in:
parent
1c9cb86381
commit
802e51430c
7 changed files with 359 additions and 0 deletions
|
|
@ -0,0 +1,56 @@
|
|||
# UNDF: UNDF-2026-000001301
|
||||
# CWE-407: Algorithmic Complexity — O(B×U) → O(B+U) in PhlareDB.GetBlockStats
|
||||
#
|
||||
# Defect: pkg/phlaredb/phlaredb.go GetBlockStats iterates three block sets
|
||||
# (heads, flushing, queriers) and for EACH block calls
|
||||
# slices.Contains(req.Msg.GetUlids(), h.meta.ULID.String()). slices.Contains
|
||||
# is O(U) linear scan. Per-request cost: O(B × U) where B = total blocks
|
||||
# across all three sets, U = requested ULID count.
|
||||
#
|
||||
# Real-world scale: pyroscope tenants storing weeks of continuous profiles
|
||||
# accumulate thousands of block queriers. Operators issuing block-stats
|
||||
# queries with hundreds of ULIDs pay 1M+ membership checks per call.
|
||||
# ULID.String() also re-formats per iteration, multiplying allocations.
|
||||
#
|
||||
# Fix: Hoist req.Msg.GetUlids() into a map[string]struct{}{} once before the
|
||||
# loops. Per-iter cost drops from O(U) to O(1). Total cost: O(B + U).
|
||||
#
|
||||
# Complexity gate (defects/pyroscope/bench/bench-pyroscope-0001.py):
|
||||
# B=2000, U=200: defective ~50ms, fixed <2ms (>=25× speedup)
|
||||
# k-scaling 5×: time ratio must be <17.5×
|
||||
--- a/pkg/phlaredb/phlaredb.go
|
||||
+++ b/pkg/phlaredb/phlaredb.go
|
||||
@@ -594,16 +594,21 @@ func (f *PhlareDB) GetBlockStats(ctx context.Context, req *connect.Request[inges
|
||||
defer sp.Finish()
|
||||
|
||||
res := &ingestv1.GetBlockStatsResponse{}
|
||||
+ // Hoist requested ULIDs into a set so per-block membership is O(1) instead of
|
||||
+ // O(U) slices.Contains. GetBlockStats walks heads + flushing + queriers
|
||||
+ // (potentially thousands of blocks); the linear scan is O(B*U) per request.
|
||||
+ requested := make(map[string]struct{}, len(req.Msg.GetUlids()))
|
||||
+ for _, u := range req.Msg.GetUlids() {
|
||||
+ requested[u] = struct{}{}
|
||||
+ }
|
||||
f.headLock.RLock()
|
||||
for _, h := range f.heads {
|
||||
- if slices.Contains(req.Msg.GetUlids(), h.meta.ULID.String()) {
|
||||
+ if _, ok := requested[h.meta.ULID.String()]; ok {
|
||||
res.BlockStats = append(res.BlockStats, h.GetMetaStats().ConvertToBlockStats())
|
||||
}
|
||||
}
|
||||
for _, h := range f.flushing {
|
||||
- if slices.Contains(req.Msg.GetUlids(), h.meta.ULID.String()) {
|
||||
+ if _, ok := requested[h.meta.ULID.String()]; ok {
|
||||
res.BlockStats = append(res.BlockStats, h.GetMetaStats().ConvertToBlockStats())
|
||||
}
|
||||
}
|
||||
f.headLock.RUnlock()
|
||||
|
||||
f.blockQuerier.queriersLock.RLock()
|
||||
for _, q := range f.blockQuerier.queriers {
|
||||
- if slices.Contains(req.Msg.GetUlids(), q.meta.ULID.String()) {
|
||||
+ if _, ok := requested[q.meta.ULID.String()]; ok {
|
||||
res.BlockStats = append(res.BlockStats, q.GetMetaStats().ConvertToBlockStats())
|
||||
}
|
||||
}
|
||||
f.blockQuerier.queriersLock.RUnlock()
|
||||
Loading…
Add table
Add a link
Reference in a new issue