roachtest: kv/contention/nodes=4 failed #45700
Assignees
Labels
branch-master
Failures and bugs on the master branch.
C-test-failure
Broken test (automatically or manually discovered).
O-roachtest
O-robot
Originated from a bot.
release-blocker
Indicates a release-blocker. Use with branch-release-2x.x label to denote which branch is blocked.
Milestone
Comments
cockroach-teamcity
added
branch-master
|
(roachtest).kv/contention/nodes=4 failed on master@c9b189b3a2171c8da78c75fe80dd02933e18a5e5: More
Artifacts: /kv/contention/nodes=4
See this test on roachdash |
|
(roachtest).kv/contention/nodes=4 failed on master@954fe69d554162aec0fbc001aad1fe5103d8df13: More
Artifacts: /kv/contention/nodes=4
See this test on roachdash |
|
(roachtest).kv/contention/nodes=4 failed on master@752dea867f3aeb142e98c22f8d320ce19041aa8d: More
Artifacts: /kv/contention/nodes=4
See this test on roachdash |
|
(roachtest).kv/contention/nodes=4 failed on master@dfa5bd527ae7d7373dd03c62118df87a87a77130: More
Artifacts: /kv/contention/nodes=4
See this test on roachdash |
|
(roachtest).kv/contention/nodes=4 failed on master@c473f40078994551cebcbe00fdbf1fa388957658: More
Artifacts: /kv/contention/nodes=4
See this test on roachdash |
|
(roachtest).kv/contention/nodes=4 failed on master@72c4a1bd411f2f82bf9aaa22883821a946614148: More
Artifacts: /kv/contention/nodes=4
See this test on roachdash |
|
(roachtest).kv/contention/nodes=4 failed on master@5570c01402796edb7cd06eb8ce7f615371f22d42: More
Artifacts: /kv/contention/nodes=4
See this test on roachdash |
|
(roachtest).kv/contention/nodes=4 failed on master@793a9200c16693aff32aa6a4dd9d8bbcbddb30aa: More
Artifacts: /kv/contention/nodes=4
See this test on roachdash |
|
(roachtest).kv/contention/nodes=4 failed on master@69dc87d68addedf2fabfb2b14c098cfb35b5f3d0: More
Artifacts: /kv/contention/nodes=4
See this test on roachdash |
nvanbenschoten
added a commit
to nvanbenschoten/cockroach
that referenced
this issue
Mar 14, 2020
Relates to cockroachdb#45700. This commit fixes a bug in the lockTableWaiter where we manually cancelled the context (to clean up resources) used by an async push _before_ checking whether the context was cancelled. This meant that we alway hit the `pushCtx.Err() == context.Canceled` path and always cleared the push error. This lead to deadlocks because a transaction would detect that it was aborted but the error would be swallowed due to this bug so the transaction would continue to wait instead of exiting its lock wait-queues. Release justification: fixes a high-severity bug where user transactions could deadlock indefinitely.
|
(roachtest).kv/contention/nodes=4 failed on master@f5585e933a097b53242d8d5800127b821a9a4d41: More
Artifacts: /kv/contention/nodes=4
See this test on roachdash |
|
(roachtest).kv/contention/nodes=4 failed on master@33d71472dc01cbc5064b3c5e1fcd666a33f606de: More
Artifacts: /kv/contention/nodes=4
See this test on roachdash |
|
(roachtest).kv/contention/nodes=4 failed on master@d8bdc938177ea4b5226b95ee6cc51af0951054fc: More
Artifacts: /kv/contention/nodes=4
See this test on roachdash |
craig bot
pushed a commit
that referenced
this issue
Mar 16, 2020
46015: sql: add ability to alter primary key of a new table r=jordanlewis a=rohany Fixes #46008. Fixes #46007. Release justification: low risk change that adds a useful feature Release note (sql change): This PR makes it possible to create a table and then add/alter primary key within the same transaction. 46058: sql: add telemetry for ORDER BY and INTERLEAVE r=yuzefovich a=yuzefovich Release justification: reporting changes only. This commit adds the reporting of telemetry on the usage of INTERLEAVE (in `createTableNode.startExec`) and of ORDER BY (by marking `opt.SortOp` with "Telemetry" flag in the optbuilder and reusing the same "infrastructure" as relational operators do, right before constructing the sort). Also it adds a logic test for the telemetry collection on the window functions' usage. Addresses: #45210. Release note: None 46059: sql: add the 'every' aggregate function r=rohany a=rohany Fixes #45842. Release justification: low risk update to functionality Release note (sql change): This PR adds the `every` aggregate function. 46099: pgwire: add telemetry for portal related requests r=rohany a=otan Resolves #43763 by covering the happy case. This PR adds telemetry everytime a "limitedCommandResult" is made, which signifies jdbc streaming. Release justification: simply telemetry. Release note: None 46101: kvserver/concurrency: check context cancellation before cancelling r=nvanbenschoten a=nvanbenschoten Relates to #45700. This commit fixes a bug in the lockTableWaiter where we manually cancelled the context (to clean up resources) used by an async push _before_ checking whether the context was cancelled. This meant that we alway hit the `pushCtx.Err() == context.Canceled` path and always cleared the push error. This lead to deadlocks because a transaction would detect that it was aborted but the error would be swallowed due to this bug so the transaction would continue to wait instead of exiting its lock wait-queues. Release justification: fixes a high-severity bug where user transactions could deadlock indefinitely. 46105: roachtest: adjust tpchvec r=yuzefovich a=yuzefovich Release justification: non-production code changes. This commit adjusts tpchvec test in the following ways: - it disables the verbose logging of `bytes_usage` (this was enabled to get something to dig into in case a fatal OOM error that occurred once would occur again, but it didn't) - it introduces two "run configurations" into the test: 1. keeps the previous behavior with a minor tweak (disables the randomization of `workmem` setting). This configuration is meant to be used to check the correctness of the vectorized engine and compare the queries' runtimes against row-by-row engine. 2. adds a new configuration in which only the vectorized engine runs all queries once, but the workmem is randomized in [100KiB, 1000KiB) range. Also, we enable logging on `vectorized_flow` file (to see that the spilling does occur). This configuration is meant to stress disk spilling of the vectorized engine. There is no comparison of the runtimes (since the row-by-row engine is not used in this config). Fixes: #46131. Release note: None 46124: sql,cli: improve the handling of notices r=otan a=knz Fixes #45833. tldr: - notices printed at end in SQL shell (requested by @RaduBerinde), - new built-in function `crdb_internal.notice()`, - notices can now be logged with vmodule `notices=2`, - code simplification. Details: The main purpose of this patch was to delay the printing of notices in the crdb SQL shell to the end, after the results. The motivation for that is that if there are many (e.g. thousands) of result rows, a notice printed at the beginning or "in the middle" would get drowned and missed by the human observer. This patch achieves that with a straightforward, local change to the CLI SQL code. To unit test the change, a way to introduce notices at an arbitrary point of query execution was also needed. To achieve this, the patch introduces a new SQL built-in function `crdb_internal.notice()` with the corresponding plumbing throughout the execution layers. This makes it possible to send an out-of-band notice in the middle of SQL execution. The unit test for the first changes already uses this new built-in; I also foresee that we will discover that our users will start using that in fancy ways that we are not envisioning yet. To troubleshoot the implementation of the above, a way to inspect the notices inside the SQL execution layer was needed; so this patch also ensures notices are logged when the vmodule config includes `notices=2`. Finally, the various changes discussed here revealed that the notice code in the SQL execution package could be simplified, so this patch does this too. Release justification: Category 4: Low risk, high reward changes to existing functionality Release note (cli change): The `sql` and `demo` client commands now display out-of-band server notices at the end of execution. 46144: sql: clean up the preliminary notices r=otan a=knz First commit from #46124. In the lead up to 20.1, we want the first impression of client notices by users to be clean and positive. Random notice messages with inconsistent casing, non-standard terminology and unscoped identifiers would likely not delight developers. To achieve this, this commit streamlines the messages for DROP INDEX and PK changes. Release justification: Category 4: Low risk, high reward changes to existing functionality Release note: None Co-authored-by: Rohan Yadav <rohany@alumni.cmu.edu> Co-authored-by: Yahor Yuzefovich <yahor@cockroachlabs.com> Co-authored-by: Oliver Tan <otan@cockroachlabs.com> Co-authored-by: Nathan VanBenschoten <nvanbenschoten@gmail.com> Co-authored-by: Raphael 'kena' Poss <knz@thaumogen.net>
|
(roachtest).kv/contention/nodes=4 failed on master@5a3d0c9539a671f0e55b680d3021b18dde9d190d: More
Artifacts: /kv/contention/nodes=4
See this test on roachdash |
|
(roachtest).kv/contention/nodes=4 failed on master@b5f030223fbcf22e806c48a3c46e74a73a54f50f: More
Artifacts: /kv/contention/nodes=4
See this test on roachdash |
|
(roachtest).kv/contention/nodes=4 failed on master@d571624ae4f833c49c717728a74cc7be78a791f0: More
Artifacts: /kv/contention/nodes=4
See this test on roachdash |
|
The |
|
(roachtest).kv/contention/nodes=4 failed on master@92495ffb2182671b3338a8c6cd2734cb6de5d392: More
Artifacts: /kv/contention/nodes=4
See this test on roachdash |
|
^ was @andreimatei can we close this issue as well as #45698 and #46461? |
|
I talked to @andreimatei about this. He saw 40 straight successful runs, so he's ok with closing. |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
(roachtest).kv/contention/nodes=4 failed on master@d6a52a8f6f07b3b6b32af2243075e5365fb21c45:
More
Artifacts: /kv/contention/nodes=4
Related:
See this test on roachdash
powered by pkg/cmd/internal/issues
The text was updated successfully, but these errors were encountered: