[MCOL-4775] DMLproc is unable to complete a rollback and goes into a crash loop Created: 2021-06-24 Updated: 2023-11-17 Resolved: 2022-06-24 |
|
| Status: | Closed |
| Project: | MariaDB ColumnStore |
| Component/s: | DMLProc |
| Affects Version/s: | 5.5.2 |
| Fix Version/s: | 6.4.1 |
| Type: | Bug | Priority: | Major |
| Reporter: | Rick Pizzi | Assignee: | Daniel Lee (Inactive) |
| Resolution: | Duplicate | Votes: | 4 |
| Labels: | None | ||
| Issue Links: |
|
||||||||||||||||||||||||||||
| Sprint: | 2021-9, 2021-10, 2021-11, 2021-12, 2021-13, 2021-14, 2021-15, 2021-16, 2021-17 | ||||||||||||||||||||||||||||
| Description |
|
A customer experienced an issue where on cluster restart, DMLproc was unable to complete the rollback of 3 tables.
You can see in the log that the process ID of DMLProc changes, but there are no traces in the log about it dying, no trace files, nothing. |
| Comments |
| Comment by Roman [ 2021-06-26 ] |
|
allen.herrera You describe unrelated but important case that should be reproduced and filed if possible. |
| Comment by Roman [ 2021-09-17 ] |
|
Agree. |
| Comment by David Hall (Inactive) [ 2022-03-04 ] |
|
Long running Rollbacks can take more than the systemd startup timeout. |
| Comment by David Hall (Inactive) [ 2022-05-13 ] |
|
We'll wait for the Pulumi project, which sets up a VM cluster for testing. Then run a test that causes failure and long rollback upon startup . |
| Comment by David Hall (Inactive) [ 2022-06-24 ] |
|
Duplicate of 5105 |