Noticed in cockroachdb#67866.

Release note: None

68156: roachtest: fix replicagc-changed-peers (again) r=erikgrinaker a=tbg

I wrote buggy code. Since it isn't hit in every invocation, it slipped
through my local testing.

Fixes #68155
Fixes #68162

Release note: None


68157: CODEOWNERS: own catalog package to obs-inf-prs r=erikgrinaker a=tbg

Noticed in #67866.

Release note: None


68203: roachtest: avoid releasing alloc multiple times r=erikgrinaker a=tbg

As of #68103, existing problems in the cluster creation retry logic
are tickled reliably: on retry we are releasing an alloc twice,
resulting in a panic. Unfortunately, the alloc is acquired several
layers above the retry, so for now the best I can do is to avoid
releasing the alloc if we are bound for a retry.

The reason the above PR tickles it more reliably is because it
accidentally removed a sensible failfast: when the cluster already
exists, we shouldn't destroy & try to recreate it again (which
would then explode on the double-release anyway).

Release note: None


68224: roachtest: fix inconsistency r=erikgrinaker a=tbg

The way that test was passing the args no longer works.

Fixes #64806.

Release note: None


68226: roachtest: allow local runs with roachstress r=erikgrinaker a=tbg

It's appealing to use roachstress also as a tool to simply run
a roachtest, since it avoids having to think much about the flags
and incantation.

This commit adds a prompt to roachstress about running in local
mode. If selected, the binaries are built to target the local
architecture, and the builder is not invoked.

Release note: None


68227: roachtest: skip acceptance/gossip/restart-node-one r=erikgrinaker a=tbg

Touches #68107.

Release note: None


Co-authored-by: Tobias Grieger <tobias.schottdorf@gmail.com>
Co-authored-by: Tobias Grieger <tobias.b.grieger@gmail.com>

…68660 #68661

67090: sql: periodically flush sqlstats r=Azhng a=Azhng

Previous PR: #67866

This commit introduces a new persisted sql stats subsystem
that wraps the existing node-local sql stats subsystem.
This new subsystem is responsible for flushing the in-meory
statistics into the system table periodically, or when it
detects memory pressure.
This replaces sql.Server's in-memory sqlStats provider.

Release note (sql change): SQL stats now can be persisted into
system.statement_statistics and system.transaction_statistics
tables by enabling the sql.stats.flush.enable cluster setting.
The interval of persistence is determined by the new
sql.stats.flush.interval cluster setting which defaults to 1 hour.

68195: streamingccl: allow stream ingestion processors to keep running on `GenerationEvent` r=annezhu98 a=annezhu98

Previously, a stream ingestion processor would shut down if it ever loses connection with its stream client. With generation support, the processor should not immediately move to draining state, instead, it should be in `StateRunning` to poll for cutover signal sent by the coordinator. Generation support will be implemented by the following PR: #67189

The first commit adds `GenerationEvent` as a valid event type that can be emitted over a cluster stream.
The second commit implements the mechanism that keeps processors running when losing connection with the client.

68288: changefeedccl: Propagate pushback throughout changefeed pipeline. r=ajwerner a=miretskiy

Propagate pushback information throughout changefeed pipeline.

The pushback propagation is accomplished by associating a `Resource`
with each event that is processed by changefeed system.  The resource is
propagated throughout changefeed, and is released when the event
has been written to the sink.

This change also updates and simplifies event memory accounting.
Prior to this PR, memory accounting was incomplete and was error prone.
This PR simplifies memory accounting by allocating resources once when
the event arrives (blocking buffer), and releasing resources
when event is written to the sink.  Dynamic modifications to the amount
of allocated memory are, in general, not safe (without additional
complexity).  To accommodate the fact that during event processing we
use more memory (e.g. to parse and convert this event), we over-allocate
resources to the the event.

Release Notes: Enterprise change; changefeed will slow down correctly
whenever there is a slow down in the system (i.e. downstream sink is
slow).

68444: multiregionccl: add zone config waiting for partitioned tables r=arulajmani a=pawalt

Previously, wait-for-zone-configs could only wait on tables, meaning we
couldn't wait for zone config changes to apply to REGIONAL BY ROW
partitions. This PR adds a `partition-name` flag to allow users to wait
on a specific partition.

Release note: None

A future PR will extend the tracing to allow us to trace REGIONAL BY ROW queries, but those changes combined with these will likely be too large for a single PR.

68596: authors: add pseudomuto to authors r=pseudomuto a=pseudomuto

Release note: None

68611: ci: move bazel build scripts to `build/teamcity` directory r=rail a=rickystewart

This gives a little more structure to what was previously a completely
flat `build` directory that contained one or more scripts for every
single build configuration.

When this PR is merged I'll move the affected build configurations to
final destinations in the `Cockroach / CI` subproject in TC.

Release note: None

68614: server: reduce flakiness of TestEnsureSQLStatsAreFlushedForTelemetry r=Azhng a=Azhng

Previously, this unit tests generates statements with the fingerprint
'INSERT INTO _ VALUES (_)'. However, this is a very common fingerprint
and can collide with actual statements issued by other subsystems, which
in turn cause test failure.
This commit change the test cases to use uncommon statements that should
not cause fingerprint collision

Resolves #66826

Release note: None

68637: kvserver/closedts: fix nil dereference in HTML generation r=aliher1911 a=erikgrinaker

The `/debug/closedts-{sender,receiver}` endpoints could panic due to a
nil dereference if the last update in the sidetransport buffer was
removed by the time it was rendered. This patch adds a `nil` check to
avoid that panic.

Release note (bug fix): Fixed a crash in the
`/debug/closedts-{sender,receiver}` advanced debug pages if the last
message of the closed timestamp side transport buffer was removed before
rendering.

68642: kvserver: deflake TestNonVoterCatchesUpViaRaftSnapshotQueue r=aayushshah15 a=aayushshah15

Fixes #68142

Release note: None

68660: authors: Add shiranka to authors r=samiskin a=samiskin



68661: add ding@ to AUTHORS r=davidwding a=davidwding



Co-authored-by: Azhng <archer.xn@gmail.com>
Co-authored-by: Anne Zhu <anne.zhu@cockroachlabs.com>
Co-authored-by: Yevgeniy Miretskiy <yevgeniy@cockroachlabs.com>
Co-authored-by: Peyton Walters <peyton.walters@cockroachlabs.com>
Co-authored-by: David Muto (pseudomuto) <david.muto@gmail.com>
Co-authored-by: Ricky Stewart <ricky@cockroachlabs.com>
Co-authored-by: Erik Grinaker <grinaker@cockroachlabs.com>
Co-authored-by: Aayush Shah <aayush.shah15@gmail.com>
Co-authored-by: Shiranka Miskin <shiranka.miskin@gmail.com>
Co-authored-by: David Ding <ding@cockroachlabs.com>