Skip to content

Bump JasperFx to 2.21.1; dynamic-lifecycle tests pin the jasperfx#492 per-tenant HWM cadence fix#4876

Merged
jeremydmiller merged 1 commit into
masterfrom
chore/jasperfx-2211-per-tenant-hwm-cadence
Jul 7, 2026
Merged

Bump JasperFx to 2.21.1; dynamic-lifecycle tests pin the jasperfx#492 per-tenant HWM cadence fix#4876
jeremydmiller merged 1 commit into
masterfrom
chore/jasperfx-2211-per-tenant-hwm-cadence

Conversation

@jeremydmiller

Copy link
Copy Markdown
Member

Part of epic #486 (WS1 residue). Consumes JasperFx 2.21.1 (#493), which drives the vectorized per-tenant high-water poll from the daemon's own SlowPollingTime timer instead of solely from store-global mark changes.

What changes

  • Directory.Packages.props: JasperFx pins 2.21.02.21.1.
  • The two dynamic-tenant-lifecycle tests (single-database + sharded flavors) previously seeded the mid-run tenant above the store's global high water — a loudly-commented workaround forcing a global tick so the new tenant's agent couldn't stall at ceiling 0 if coordinator discovery won the race against the append. They now seed the new tenant below the global mark, which deletes the workaround and turns both tests into active regression pins: against 2.21.0 this exact shape stalls the new tenant's agent indefinitely.

For scale-level evidence of the same stall (65/100 tenants each exactly one append-batch short of ceiling, 100/100 after the fix) see the daemonload scenario in #4875 and the epic comment on #486.

Verification

Full TenantPartitionedEventsTests suite green locally against 2.21.1 (215 tests, net9.0, Postgres 5432).

🤖 Generated with Claude Code

… global mark, pinning #492

JasperFx 2.21.1 (#493) drives the vectorized per-tenant high-water
poll from the daemon's own SlowPollingTime timer instead of solely from
store-global mark changes, so a tenant whose appends stay below the global
max converges without needing unrelated activity.

The two dynamic-tenant-lifecycle tests (single-database + sharded flavors)
previously worked around the cadence gap by seeding the mid-run tenant ABOVE
the store's global high water to force a global tick on either side of the
discovery-vs-append race. They now seed the new tenant BELOW the global mark
— which both deletes the workaround and turns the tests into active
regression pins for the fix: against 2.21.0 this exact shape stalled the new
tenant's agent at 0 indefinitely.

Full TenantPartitionedEventsTests suite green (215 tests, net9.0).

Part of epic #486 (WS1 residue).

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
@jeremydmiller
jeremydmiller merged commit 769cd23 into master Jul 7, 2026
9 checks passed
@jeremydmiller
jeremydmiller deleted the chore/jasperfx-2211-per-tenant-hwm-cadence branch July 7, 2026 12:34
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