|
Looks like MDEV-9510 which has a fix coming in 10.1.30
|
|
Closing as it's presumably fixed in the upcoming 10.1.30 release, in the scope of MDEV-9510. If you keep experiencing the problem after upgrading to 10.1.30 or higher, please comment and the report will be re-opened.
|
|
elenst, The backtrace from mariadb4 looks like something new and different from MDEV-9510. I coudn't see anything like it but could you check too please.
|
|
danblack, there are examples of this kind, too, in the comments to MDEV-9510.
anikitin, would you like to triple-check that it's the same issue as MDEV-9510 (in all provided variations)?
|
|
The crashes in mark_xid_done definitely look like MDEV-9510; do_checkpoint_request still calls mark_xid_done explicitly : https://github.com/MariaDB/server/blob/1c4968f2f39ee7796853d70d71b061989293bacb/sql/log.cc#L6555 , so I don't see any sign why it can't be MDEV-9510. Additionally, the fix for MDEV-9510 is still "best effort", meaning that there is no logical guarantee that no similar crashes will happen. I don't see any way to prove or disprove that it is or isn't MDEV-9510. So I suggest to wait and see if anybody claims similar problems in version where 9510 is fixed.
|