mirror of
https://github.com/hanzoai/datastore.git
synced 2026-09-26 13:55:07 +00:00
Under sanitizers the per-flush symbolization (SymbolIndex::findSymbol + tryDemangle + AddressToLineCache) cannot keep up with the sample rate from the global + per-query profilers. Single flushes were observed taking up to 2.5 minutes for ~180k entries, blocking any test that issues SYSTEM FLUSH LOGS. Symbols are still resolvable on read via addressToSymbol/addressToLine, so tests that need them are unaffected. CI report: https://s3.amazonaws.com/clickhouse-test-reports/json.html?PR=100242&sha=35fcb7327e723b2670a12c8e03723b0fec7bdeaf&name_0=PR&name_1=Stateless%20tests%20%28amd_tsan%2C%20parallel%2C%201%2F2%29 PR: https://github.com/ClickHouse/ClickHouse/pull/100242 Co-Authored-By: Claude Opus 4.7 (1M context) <[email protected]>
11 lines
517 B
XML
11 lines
517 B
XML
<clickhouse>
|
|
<!-- Under sanitizer builds, in-flush symbolization (SymbolIndex::findSymbol +
|
|
tryDemangle + AddressToLineCache) is too slow to keep up with the sample rate
|
|
from the global + per-query profilers. The trace_log flush thread falls
|
|
behind, backlog grows, and tests that issue SYSTEM FLUSH LOGS can wait
|
|
minutes. Resolve symbols on read instead via addressToSymbol/addressToLine. -->
|
|
<trace_log>
|
|
<symbolize>false</symbolize>
|
|
</trace_log>
|
|
</clickhouse>
|