Plans for v3.4.20 release

[ahrtr]

We intentionally skipped back-porting some bug fixes in 3.4.19 in order to minimize the impact. Since 3.4.19 is already released and validated to be working well, so we can consider to release 3.4.20 to backport some PRs.

Pipelines/test

There are still some flaky test failures from time to time. As previous release, any fix to stabilize the pipeline is welcome.

• [3.4] Test failure for TestRoundRobinBalancedResolvableFailoverFromServerFail #14216
  • [3.4] clientv3/balancer: fixed flaky TestRoundRobinBalancedResolvableFailoverFromServerFail #14274
• [3.4] panic: runtime error: invalid memory address or nil pointer dereference #14256
  • [3.4] Do not get previous K/V for create event #14290
• [3.4] ./test: line 229: kill: (7520) - No such process #14257
• [3.4] panic: runtime error: index out of range [0] with length 0 [recovered] #14259
• [3.4] WARNING: DATA RACE between mvcc.(*store).setupMetricsReporter and mvcc.(*store).restore #14270
  • [3.4] clientv3/mvcc: fixed DATA RACE #14279

Issues/PRs that should be included in 3.4.20

• raft: postpone MsgReadIndex until first commit in the term #12762
  • [3.4] raft: postpone MsgReadIndex until first commit in the term #14258
• [3.4] etcdserver: resend ReadIndex request on empty apply request #14269
• etcdserver: remove temp files in snap dir when etcdserver starting #12846
  • [3.4] etcdserver,pkg: remove temp files in snap dir when etcdserver starting #14246
• [3.5] Support linearizable renew lease #13932
  • [3.4] Support linearizable renew lease for 3.4 #14177
• [release-3.5] clientv3: filter learners members during autosync #13917
  • [release-3.4] clientv3: filter learners members during autosync #14236
• backport 3.5: #13676 load all leases from backend #13726
  • [3.4] backport 3.5: #13676 load all leases from backend #14239
• *: fix IsOptsWithFromKey #13736
  • [3.4] clientv3: fix isOptsWithFromKey/isOptsWithPrefix #14241
• Lease Checkpoints fix #13508
  • [3.4] Checkpoints fix 3.4 #14253
• [3.4] server/auth: protect rangePermCache with a RW lock #14230
• etcdserver: add more detailed traces on linearized reading #12335
  • [3.4] etcdserver: add more detailed traces on linearized reading #14240
• Support configuring MaxConcurrentStreams for http2 #14169
  • [3.4] Support configuring MaxConcurrentStreams for http2 #14251
• etcdserver: non-mutating requests pass through quotaKVServer when NOS… #13435
  • [3.4] Backport: non mutating requests pass through quotaKVServer when NOSPACE #14254

Issues/PRs not considered in 3.4.20

• etcdserver: significantly reduces start-up time #11779
  See etcdserver: significantly reduces start-up time #11779 (comment)

Please feel free to chime in if you think any PR/issues need to be investigated or included in 3.4.20. Please also feel free to let us know if anyone has any concerns or comment. cc @ptabor @serathius @spzala @hexfusion @mitake @lavacat @endocrimes @dims @liggitt @neolit123

[chaochn47]

Nice to backport: #12335

[ahrtr]

Nice to backport: #12335

Thanks for the feedback. I just had a quick review on the PR, and it's a minor change and should be safe to backport. Please feel free to deliver a PR for this.

[vivekpatani]

@ahrtr #14241

[tangcong]

@ahrtr #14169

[ahrtr]

@ahrtr #14169

Thanks for the feedback. I agree the PR can improve the performance/throughput, but it also added a new flag --max-concurrent-streams which usually we shouldn't backport. We need to get consensus on this one. Please kindly provide your opinion (thumb up or down?). thx.

Please read #14169 (comment) if you want to quickly understand the context of the PR.

Personally, I incline to backporting it to 3.4.