Skip to content

test: give ClusterQuad the data_port_span reservation#37

Merged
sqlrush merged 1 commit into
mainfrom
fix/73-quad-span
Jul 11, 2026
Merged

test: give ClusterQuad the data_port_span reservation#37
sqlrush merged 1 commit into
mainfrom
fix/73-quad-span

Conversation

@sqlrush

@sqlrush sqlrush commented Jul 11, 2026

Copy link
Copy Markdown
Owner

Completes the spec-7.3 span root-fix: ClusterQuad was missed (Pair/Triple covered), so quad rigs cross-wire worker listeners under the shipped default pool -- t/327 died at boot/formation in nightly 29129633912 and 29138202141, bailing the stage5-integrated shard and masking t/328-331. Mirror of the Triple implementation; t/327 locally 5/5 green.

Test plan:

  • t/327 locally green
  • fast-gate
  • re-tag nightly after merge (expect integrated shard to progress to its known-red t/328 leg or turn green)

Spec: spec-7.3-lms-worker-pool.md

The spec-7.3 span root-fix covered ClusterPair and ClusterTriple but
not ClusterQuad, which still reserves ONE data port per node -- under
the shipped default pool of 2 the quad's worker listeners cross-wire
consecutive free ports (racy: boot death or formation timeout,
whichever port the allocator lands on).  t/327 died both ways across
nightly runs 29129633912 / 29138202141, bailing out the
stage5-integrated-acceptance shard and masking t/328-331.

Mirror the Triple implementation: reserve [data_port, data_port+span)
per node, span default 2 following cluster.lms_workers = 2; six quad
rigs (t/326/327/331/343/344/362) get the deterministic reservation.

Verified locally: t/327 5/5 green.

Spec: spec-7.3-lms-worker-pool.md
@sqlrush sqlrush merged commit b015369 into main Jul 11, 2026
5 checks passed
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant