Record loaded-host upload latency during a 50000-file User purge #1476

Open
opened 2026-10-08 16:42:52 +00:00 by kayg · 0 comments
Owner

Merge35 triage (#1458) ran the remaining User deletion phase on the real merged release. The 50,000-file Home purge retained all 20 parallel uploads byte for byte; all uploads returned HTTP 201. It exceeded the local latency target. These are latency-only findings, not a merge blocker under the owner's SLOW-only rule.

Local shared-host measurements, eight burst workers, load average [43.3599, 26.8628, 21.0283]:

Measurement Seconds
Ordinary uploads, p50 / p95 0.384 / 3.319
Uploads during purge, p50 / p95 34.034 / 54.894
Purge duration 78.727

Server CPU rose by 3.45 seconds; RSS changed from 285,687,808 to 357,023,744 bytes. docs/perf/baseline.json has no matching User purge profile. The probe records #1420's prior local p95 of 23.069 seconds, which is not a quiet-host controlled baseline. Do not infer an idle-host regression from this loaded run. Recheck on the perf VM through perf-run with a recorded source revision and load.

The original 2-second assertions remain unchanged. The probe logs these under its ordinary findings rather than its SLOW list; classification here is based on the completed HTTP results and exact-byte controls. Evidence: artifacts/m35triage/remaining.log, "User purge profile". Binary SHA-256: 3a6f4dd08b52923519613f2a91853dab3a3bd40e39f3c71f52b18c0c9f3298d7. The profile's revision field was unspecified; this binary identity supplies the source control. No performance measurement is claimed for the perf VM. No push or deploy.

Merge35 triage (#1458) ran the remaining User deletion phase on the real merged release. The 50,000-file Home purge retained all 20 parallel uploads byte for byte; all uploads returned HTTP 201. It exceeded the local latency target. These are latency-only findings, not a merge blocker under the owner's SLOW-only rule. Local shared-host measurements, eight burst workers, load average [43.3599, 26.8628, 21.0283]: | Measurement | Seconds | |---|---:| | Ordinary uploads, p50 / p95 | 0.384 / 3.319 | | Uploads during purge, p50 / p95 | 34.034 / 54.894 | | Purge duration | 78.727 | Server CPU rose by 3.45 seconds; RSS changed from 285,687,808 to 357,023,744 bytes. docs/perf/baseline.json has no matching User purge profile. The probe records #1420's prior local p95 of 23.069 seconds, which is not a quiet-host controlled baseline. Do not infer an idle-host regression from this loaded run. Recheck on the perf VM through perf-run with a recorded source revision and load. The original 2-second assertions remain unchanged. The probe logs these under its ordinary findings rather than its SLOW list; classification here is based on the completed HTTP results and exact-byte controls. Evidence: artifacts/m35triage/remaining.log, "User purge profile". Binary SHA-256: 3a6f4dd08b52923519613f2a91853dab3a3bd40e39f3c71f52b18c0c9f3298d7. The profile's revision field was unspecified; this binary identity supplies the source control. No performance measurement is claimed for the perf VM. No push or deploy.
Commenting is not possible because the repository is archived.
No labels
No milestone
No project
No assignees
1 participant
Due date
The due date is invalid or out of range. Please use the format "yyyy-mm-dd".

No due date set.

Dependencies

No dependencies set

Reference
kayg/calternal-v0#1476
No description provided.