Segmentation fault during rolling update

[tc-hsteffen]

I am running 27 galeraclusters on Kubernetes. Sporadically, I see an issue during rolling updates, which leads to a full SST, which I would rather avoid.

I posted a bugreport at mariadb a year ago, but it hasn't been considered yet. Maybe mariadb is the wrong addressee after all, and it's more a question to the codership galera team?

The full bug report can be found here:
https://jira.mariadb.org/browse/MDEV-33518

Thank you

[janlindstrom]

For some reason that MDEV is not assigned to anyone and in Major priority so it has been on queue. Do you have better error log with stacktrace? MariaDB version reported is quite old so does your problem reproduce with more recent version?

[tc-hsteffen]

Thank you for getting back to me - unfortunately, logs really stops after that line "If you see no messages after this, something went terribly wrong..." so I can't provide more detailed logs or stacktrace right now. I haven't tested with a more recent mariadb version yet. I'll try to reproduce it in a staging environment with a more recent version.

I am attaching a log file from a more recent crash. It looks, as if the IST completes successfully, but then, it crashes, and after the crash a full SST is started.

Do you think updating to 10.6.21 could resolve the issue? Or should we rather go directly for 11.7.2?

Explore-logs-2025-02-24 13_49_45.txt

[janlindstrom]

Hi, New log does not help more unfortunately as the stack trace is still missing. I would first try 10.6.21 as it contains all relevant fixes in this area but you also need a new Galera library version. 11.7.2 contains just new features but not any new fixes on this area. R: Jan

…

On Mon, Feb 24, 2025 at 2:58 PM Henrik Steffen ***@***.***> wrote: Thank you for getting back to me - unfortunately, logs really stops after that line "If you see no messages after this, something went terribly wrong..." so I can't provide more detailed logs or stacktrace right now. I haven't tested with a more recent mariadb version yet. I'll try to reproduce it in a staging environment with a more recent version. I am attaching a log file from a more recent crash. It looks, as if the IST completes successfully, but then, it crashes, and after the crash a full SST is started. Do you think updating to 10.6.21 could resolve the issue? Or should we rather go directly for 11.7.2? Explore-logs-2025-02-24 13_49_45.txt <https://github.com/user-attachments/files/18942613/Explore-logs-2025-02-24.13_49_45.txt> — Reply to this email directly, view it on GitHub <#673 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/ABQLREMM6V57LQ4CTVZQ75L2RMJQ7AVCNFSM6AAAAABXXX6K3KVHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMZDMNZYGM2DMNJXG4> . You are receiving this because you were assigned.Message ID: ***@***.***> [image: tc-hsteffen]*tc-hsteffen* left a comment (codership/galera#673) <#673 (comment)> Thank you for getting back to me - unfortunately, logs really stops after that line "If you see no messages after this, something went terribly wrong..." so I can't provide more detailed logs or stacktrace right now. I haven't tested with a more recent mariadb version yet. I'll try to reproduce it in a staging environment with a more recent version. I am attaching a log file from a more recent crash. It looks, as if the IST completes successfully, but then, it crashes, and after the crash a full SST is started. Do you think updating to 10.6.21 could resolve the issue? Or should we rather go directly for 11.7.2? Explore-logs-2025-02-24 13_49_45.txt <https://github.com/user-attachments/files/18942613/Explore-logs-2025-02-24.13_49_45.txt> — Reply to this email directly, view it on GitHub <#673 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/ABQLREMM6V57LQ4CTVZQ75L2RMJQ7AVCNFSM6AAAAABXXX6K3KVHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMZDMNZYGM2DMNJXG4> . You are receiving this because you were assigned.Message ID: ***@***.***>