10.5.12-MariaDB-cll-lve_rem0rec.inl
[MDEV-30787] Assertion failure in file rem0rec.inl line 375 Created: 2023-03-04 Updated: 2023-04-03 Resolved: 2023-04-03 |
|
| Status: | Closed |
| Project: | MariaDB Server |
| Component/s: | Storage Engine - InnoDB |
| Affects Version/s: | 10.5.12, 10.6.12 |
| Fix Version/s: | N/A |
| Type: | Bug | Priority: | Major |
| Reporter: | Teodoras Celencevičius | Assignee: | Unassigned |
| Resolution: | Incomplete | Votes: | 2 |
| Labels: | None | ||
| Environment: |
Cloudlinux 7.9 4.18.0-305.19.1.lve.el7h.x86_64 |
||
| Attachments: |
10.5.12-MariaDB-cll-lve_rem0rec.inl
|
| Description |
|
MariaDB server instance suddenly started to fail with error signal 6:
The server only started with innodb_force_recovery = 3. The server was then upgraded to 10.6.12-MariaDB-cll-lve, and the issue persisted. This was only solved by dropping all databases and reimporting them. Attaching related logs + backtrace. |
| Comments |
| Comment by Marko Mäkelä [ 2023-03-04 ] |
|
The assertion fails in the function rec_get_n_fields() because a page is corrupted. Can you please provide a fully resolved stack trace of the crash, with the appropriate -debuginfo or -dbgsym package installed? The stack trace in 10.5.12-MariaDB-cll-lve_rem0rec.inl |
| Comment by Paulius mickus [ 2023-03-04 ] |
|
I'm afraid it's not possible to provide a fully resolved stack trace since we've re-installed MariaDB server and re-imported all the databases from dumps meaning we're unable to reproduce this anymore. We're going to follow provided guide and install appropriate packages to be able provide full strack trace in case this issue happens again in the future. |
| Comment by Marko Mäkelä [ 2023-03-05 ] |
|
I wonder why the page passed checksum checks. Was it perhaps corrupted while the page was loaded into the buffer pool, and then written back to the file with a correct checksum. A copy of the corrupted page would be very helpful. |
