sql: accept WITH (WAIT ...) on ALTER CLUSTER unconditionally - #38104
Closed
aljoscha wants to merge 8 commits into
Closed
sql: accept WITH (WAIT ...) on ALTER CLUSTER unconditionally#38104aljoscha wants to merge 8 commits into
aljoscha wants to merge 8 commits into
Conversation
This was referenced Aug 7, 2026
aljoscha
force-pushed
the
aljoscha/cluster-legacy-04-zdt-flag
branch
from
August 7, 2026 08:00
b3e6c5d to
44e8e16
Compare
aljoscha
force-pushed
the
aljoscha/cluster-legacy-03-staged-machine
branch
from
August 7, 2026 08:01
385e39c to
281b832
Compare
aljoscha
force-pushed
the
aljoscha/cluster-legacy-04-zdt-flag
branch
from
August 7, 2026 08:01
44e8e16 to
e06b9d6
Compare
aljoscha
marked this pull request as ready for review
August 7, 2026 08:14
aljoscha
force-pushed
the
aljoscha/cluster-legacy-04-zdt-flag
branch
from
August 7, 2026 08:28
e06b9d6 to
8f88cec
Compare
aljoscha
force-pushed
the
aljoscha/cluster-legacy-03-staged-machine
branch
from
August 7, 2026 08:28
281b832 to
6aacdb9
Compare
aljoscha
force-pushed
the
aljoscha/cluster-legacy-04-zdt-flag
branch
from
August 7, 2026 08:45
8f88cec to
f1f08b2
Compare
aljoscha
force-pushed
the
aljoscha/cluster-legacy-03-staged-machine
branch
from
August 7, 2026 08:45
6aacdb9 to
0c59e5b
Compare
aljoscha
force-pushed
the
aljoscha/cluster-legacy-04-zdt-flag
branch
from
August 7, 2026 10:05
f1f08b2 to
c6c719e
Compare
5 tasks
aljoscha
force-pushed
the
aljoscha/cluster-legacy-03-staged-machine
branch
from
August 10, 2026 17:52
e115a22 to
86f1766
Compare
aljoscha
force-pushed
the
aljoscha/cluster-legacy-04-zdt-flag
branch
from
August 10, 2026 17:52
c6c719e to
d9f10e3
Compare
aljoscha
force-pushed
the
aljoscha/cluster-legacy-04-zdt-flag
branch
from
August 17, 2026 09:58
d9f10e3 to
cc7aa1f
Compare
aljoscha
force-pushed
the
aljoscha/cluster-legacy-03-staged-machine
branch
from
August 17, 2026 09:58
86f1766 to
fbc61b7
Compare
Contributor
|
Fine to remove the "private preview" status from my side! cc @maheshwarip |
ggevay
approved these changes
Aug 23, 2026
aljoscha
force-pushed
the
aljoscha/cluster-legacy-04-zdt-flag
branch
from
August 24, 2026 09:39
cc7aa1f to
93798f7
Compare
aljoscha
force-pushed
the
aljoscha/cluster-legacy-03-staged-machine
branch
from
August 24, 2026 09:39
fbc61b7 to
958d547
Compare
aljoscha
force-pushed
the
aljoscha/cluster-legacy-04-zdt-flag
branch
from
August 24, 2026 09:44
93798f7 to
7328043
Compare
aljoscha
force-pushed
the
aljoscha/cluster-legacy-03-staged-machine
branch
from
August 24, 2026 09:44
958d547 to
cb3f460
Compare
aljoscha
force-pushed
the
aljoscha/cluster-legacy-04-zdt-flag
branch
from
August 24, 2026 09:46
7328043 to
cac31da
Compare
aljoscha
force-pushed
the
aljoscha/cluster-legacy-04-zdt-flag
branch
2 times, most recently
from
August 31, 2026 12:23
a3ff998 to
c2743f3
Compare
aljoscha
force-pushed
the
aljoscha/cluster-legacy-03-staged-machine
branch
2 times, most recently
from
August 31, 2026 14:30
c6f5a4c to
fe8f383
Compare
aljoscha
force-pushed
the
aljoscha/cluster-legacy-04-zdt-flag
branch
from
August 31, 2026 14:30
c2743f3 to
6b0fdfd
Compare
aljoscha
force-pushed
the
aljoscha/cluster-legacy-03-staged-machine
branch
from
August 31, 2026 14:40
fe8f383 to
3fc3358
Compare
aljoscha
force-pushed
the
aljoscha/cluster-legacy-04-zdt-flag
branch
from
August 31, 2026 14:40
6b0fdfd to
02e2ef7
Compare
Contributor
Author
|
Fixed. Current upstream main is |
Remove ALTER-time prediction of the controller's transient strategy union. Concrete replica-create transactions enforce count and credit limits, and an exhausted graceful reconfiguration keeps serving from its realized replicas while recording a resource-exhausted status and audit transition. Keep only intrinsic target-baseline and replication-factor checks in the sequencer. A forced cut-over waits for its complete target set to exist, while the baseline yields the realized set once the commit deadline has passed. The controller creates the target and retires the realized replicas in one catalog transaction, so a zero-wait replacement needs room for the settled target rather than both sets. Catalog implications queue same-cluster drops before creates so physical orchestration quotas honor the same net accounting contract. Derive create-time scoped configuration from concrete catalog create ops at the coordinator transaction boundary. This makes controller-created replacement replicas receive render-frozen overrides before their first render, regardless of which component produced the create. Foreground waits report matching resource exhaustion as insufficient resources instead of a misleading timeout. A later ALTER that supersedes the waiter's target cannot misattribute its own failure.
aljoscha
force-pushed
the
aljoscha/cluster-legacy-04-zdt-flag
branch
from
August 31, 2026 15:15
02e2ef7 to
4dd05f1
Compare
aljoscha
force-pushed
the
aljoscha/cluster-legacy-03-staged-machine
branch
2 times, most recently
from
August 31, 2026 15:32
3798c93 to
a850316
Compare
aljoscha
force-pushed
the
aljoscha/cluster-legacy-04-zdt-flag
branch
from
August 31, 2026 15:32
4dd05f1 to
3506904
Compare
Preserve the design's historical framing while recording the few implementation refinements readers need to interpret it accurately.
The controller owns every managed cluster's replica set. Shape-changing ALTERs return through the durable reconfiguration-record path before reaching the managed-to-managed config write. The `controller_owns = true` guard therefore left the staged producer and sequencer reconciliation branches unreachable. Deleting those branches does not change managed-cluster behavior. Delete the WaitForHydrated and Finalize stages, pending overlap replicas, per-connection pending ALTER state, sequencer replica create/drop branches, and their dead op and error plumbing. Keep the durable pending field and catalog-open cleanup so a replica stranded by an older binary is still reaped after upgrade. The surviving managed-to-managed path is a config-only write. The cluster controller reconciles replication-factor changes and scheduled-cluster shape changes on its next tick.
Record the narrow compatibility reason the durable pending field and catalog-open cleanup remain even though the controller does not use pending replicas as active lifecycle state.
Graceful cluster reconfiguration has been behind the `enable_zero_downtime_cluster_reconfiguration` feature flag, default off, so the `WITH (WAIT ...)` surface is rejected at plan time unless a deployment turns it on. The controller record is now the only managed-cluster reshape path, which leaves the flag controlling only whether users can express its deadline and timeout behavior. Remove the flag and planner gate. The two rejections that share the code path stay: a `WAIT` without a replica-shape change, and a `WAIT` on an unmanaged cluster. Every test that used the surface enabled the flag itself, so those statements go. In a mixed-version run some phases execute against a released binary that still enforces the gate, so `get_minimal_system_parameters` pins the flag on below v26.41 instead. The docs drop the private-preview badges because the surface is generally available once nothing gates it.
aljoscha
force-pushed
the
aljoscha/cluster-legacy-03-staged-machine
branch
from
August 31, 2026 17:52
a850316 to
bc37a9e
Compare
aljoscha
force-pushed
the
aljoscha/cluster-legacy-04-zdt-flag
branch
from
August 31, 2026 17:52
3506904 to
0a1280c
Compare
aljoscha
force-pushed
the
aljoscha/cluster-legacy-03-staged-machine
branch
from
September 4, 2026 10:20
bc37a9e to
a0aa6db
Compare
Contributor
Author
|
Superseded by #38627, which landed the same unconditional |
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
4 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Motivation
WITH (WAIT ...)onALTER CLUSTERis gated behindenable_zero_downtime_cluster_reconfiguration, default off. Stacked on #38103,every graceful reconfiguration already runs through one durable controller path,
so the flag only controls whether users can express a deadline and a timeout
action for it.
Description
Removes the feature flag and its planner gate, so
WITH (WAIT ...)is acceptedon every deployment. The two rejections that share that code path stay: a
WAITwithout a replica-shape change, and a
WAITon an unmanaged cluster.The mixed-version default is bounded at v26.41 so upgrade scenarios running
against an older binary still set the flag, and the private-preview badges come
off the
ALTER CLUSTERreference page.enable_cluster_schedule_refresh, the near-identical sibling gate, isdeliberately untouched. It gates a separate SQL surface under its own rollout.
Verification
Existing graceful-reconfiguration coverage in testdrive, sqllogictest, and
platform checks now exercises the surface without setting a flag.
User-visible behavior
Graceful cluster reconfiguration with
WITH (WAIT UNTIL READY ...)andWITH (WAIT FOR ...)becomes generally available, no longer private preview andno longer requiring a feature flag.