Improve backtrace catching on server failures in CI for stress tests#20462
Conversation
|
#19524 does not handle SIGABRT deliberately. |
|
I see the motivation here: #19580 (comment), let's try... |
025a217 to
42f7ca6
Compare
|
This is also required in other places (like fast-test - https://clickhouse-test-reports.s3.yandex.net/20477/8ad20fcee5aaa642c2a2dd873d02103692d554f4/fast_test.html#fail1) |
42f7ca6 to
8c9b507
Compare
|
Note that it should not prevent receiving fatal messages with stack trace in client. |
Right. Actually I don't think that this should break this, since the server will continue after gdb will print stacktraces |
8c9b507 to
4b92b31
Compare
|
Stress tests are okay, but fasttest not: By some reason fasttest image wasn't updated, because that layer was in cache https://clickhouse-test-reports.s3.yandex.net/20462/014c11af7fcaa60d62c10b1497bbdc02526d180c/push_to_dockerhub/taglatest.out.1.log |
|
72c61da does not helps with avoiding using docker cache for "new" layers, maybe it should be reset somehow? (@alesapin ?) |
|
Lack of backtrace in stress tests happens one more time - https://clickhouse-test-reports.s3.yandex.net/20477/18a147b3179e5931d89395d5e46b4d833d4bef05/stress_test_(memory).html#fail1 |
Let's try install it as a separate layer then - c879719e7c1ed188cdd81a73148575cdbbacaf13 |
Still no luck, moved into separate PR - #20864 |
… same user This will allow to attach with gdb for better diagnosis.
Otherwise sometimes stracktraces may be lost [1]: [1]: https://clickhouse-test-reports.s3.yandex.net/19580/6aecb62416ece880cbb8ee3a803e14d841388dde/stress_test_(thread).html#fail1
c879719 to
770c340
Compare
Looks hanged @alexey-milovidov can you merge this? (maybe it will reproduce the problem and will have full stacktrace, in the meantime I will look into this #16481 (comment)) |
Changelog category (leave one):
Follow-up for: #19524
Cc: @alexey-milovidov