Ship Keldra 0.16.1 with three-node recovery fixes - #265

Open

zcourts wants to merge 79 commits into

mainfrom

ops/three-node-qualification

Commits on Sep 2, 2026

Retry atomic finalization after reference backlog
zcourts
committed
8addc58
Preserve authorization handoff destination errors
zcourts
committed
cd035de
Coalesce authorization handoff frames
zcourts
committed
98357f1
Make stale journal settlement idempotent
zcourts
committed
c20f475
Restore typed JSON scalar normalization
zcourts
committed
f7fdd2a
Wait for object routing after cluster growth
zcourts
committed
d398005
Run growth convergence retries outside command substitution
zcourts
committed
8b52e1b
Capture transient growth probe failures without errexit
zcourts
committed
6754a87
Qualify three-node payloads through node outages
zcourts
committed
fe5cd6e
Enable journal evidence in cluster qualification
zcourts
committed
d6e3a0a
Bound three-node journal pressure qualification
zcourts
committed
4d957b4
Align journal pressure with atomic batch bounds
zcourts
committed
8f01b8d
Use bulk writes for cluster cutover pressure
zcourts
committed
72d5b8b
Batch distributed mutation reconciliation
zcourts
committed
69b7b5a
Deduplicate distributed payload work
zcourts
committed
a976e05
Reconcile local retention during node handoff
zcourts
committed
d52fe21
Ignore replica-local retention progress in quorum reads
zcourts
committed
4894050
Propagate BulkWrite deadlines across peer routes
zcourts
committed
9c393be
Allow bounded time for pressure-tested node joins
zcourts
committed
025b682
Budget pressure joins for complete typed handoff
zcourts
committed
8594c34

Commits on Sep 3, 2026

Wait for retained reference progress during handoff
zcourts
committed
e36a3e0
Prepare Keldra 0.16.1 release
zcourts
committed
e0685df
Restore Store source size gate
zcourts
committed
17d7113
Unblock reference delivery at journal capacity
zcourts
committed
5778069
Fix three-node qualification evidence suffix
zcourts
committed
6bc44b1
Wait for quiescent journal before cutover
zcourts
committed
fe7902c
Fix learner joins behind purged Raft logs
zcourts
committed
ad8842e
Fail over credential verification within replica quorum
zcourts
committed
301c7f8
Fail over logical authority reads within replica quorum
zcourts
committed
6ad1c1c
Fail over fresh authorization reads within replica quorum
zcourts
committed
bfabc28
Allow selected replicas to reconcile authorization realms
zcourts
committed
4332315
Prefer local replicas for credential failover
zcourts
committed
bfb06c1
Isolate logical record reconciliation stalls
zcourts
committed
3b06e09
Bound logical replica outage recovery
zcourts
committed
60871f2
Bound and redrive learner catch-up
zcourts
committed
be76cc0
Redrive learners without blocking replication
zcourts
committed
4449fbd
Enforce learner progress timeout
zcourts
committed
a33427a
Serve joining nodes during background handoff
zcourts
committed
747bb4d
Allow joining coordinators before local bootstrap handoff
zcourts
committed
f310f10
Merge remote-tracking branch 'origin/main' into ops/three-node-qualification
zcourts
committed
f71bcce
Keep join handoff replay on active sources
zcourts
committed
abb6ef0
Fix background join qualification polling
zcourts
committed
6e2c678
Wait for joined cluster state before serving
zcourts
committed
b069371
Proxy list reads through joining nodes
zcourts
committed
ac87f1d
Allow joining coordinators to fetch payload reads
zcourts
committed
e663c3f
Keep payload read authorization change compact
zcourts
committed
d41b552

Commits on Sep 4, 2026

Allow joining nodes to proxy object requests
zcourts
committed
1ffd90a
Refresh Zanzibar 0.16.1 release candidate
zcourts
committed
1878cbd
Route credential exchange through joining nodes
zcourts
committed
610b88c
Route logical name reads through joining gateways
zcourts
committed
a073076
Route credential target selection through joining gateways
zcourts
committed
732c77b
Allow joining gateways to perform quorum reads
zcourts
committed
69d86ac
Fence fresh authorization reads by placement
zcourts
committed
f2a9441
Keep logical quorum reads available during join
zcourts
committed
bd8ade8
Wait for replicated durability during node join
zcourts
committed
b769bd1
Keep object link source within release limit
zcourts
committed
e0e07de
Wait for serving fence after node activation
zcourts
committed
6c99184
Route joining gateway mutations by stable identity
zcourts
committed
f15e2e5
Retry fresh authorization across placement changes
zcourts
committed
342b1f4
Select current capabilities for fresh clusters
zcourts
committed
3866b8b
Retry routed metadata reads after placement changes
zcourts
committed
0ea9382
Retry atomic program routing after executor changes
zcourts
committed
c6b5eca
Document Keldra architecture and join lifecycle
zcourts
committed
bd841f1
Fix merged TypedJson physical field IDs
zcourts
committed
78ba1db
Align release checks with fresh capability 2 clusters
zcourts
committed
0239a3e
Retry local atomic execution after executor changes
zcourts
committed
7315cc8
Retry transient atomic executor lookup
zcourts
committed
b671e7a
Validate routed replay on the destination executor
zcourts
committed
ab406f7
Allow joining nodes to proxy sealed uploads
zcourts
committed
d98b793
Retry credential exchange after placement changes
zcourts
committed
26ca65e
Accept payload evidence from joining uploaders
zcourts
committed
b361c20
Capture cutover baseline after joining probe
zcourts
committed
125786c
Grow v6 query memory with retained data
zcourts
committed
466bf81
Allow joining nodes to read transfer state
zcourts
committed
d96aff9
Rebuild v6 families after catalog changes
zcourts
committed
d00d98a
Return distributed reads after quorum
zcourts
committed
39987af
Allow selected replicas to serve logical reads
zcourts
committed
28b2b4e

Commits on Sep 5, 2026

Keep atomic executor with Raft leader
zcourts
committed
4f09238
Allow built-in distributed path stages
zcourts
committed
ea22cb8