[MDEV-16827] UNIX_TIMESTAMP() not working properly during replication Created: 2018-07-25 Updated: 2018-08-03 Resolved: 2018-08-03 |
|
| Status: | Closed |
| Project: | MariaDB Server |
| Component/s: | Replication |
| Affects Version/s: | 10.0.34-galera |
| Fix Version/s: | N/A |
| Type: | Bug | Priority: | Blocker |
| Reporter: | Brian Kolaci | Assignee: | Unassigned |
| Resolution: | Not a Bug | Votes: | 0 |
| Labels: | need_feedback | ||
| Environment: |
Replicating from MySQL 5.0.75 to MariaDB 10.0.34 Linux diskstation2 3.10.102 #15266 SMP Mon Mar 26 15:08:11 CST 2018 x86_64 GNU/Linux synology_x86_1511+ |
||
| Description |
|
I've enabled replication on a Synology diskstation. Starting replication I get:
which indicates UNIX_TIMESTAMP(NOW()+2000000)) is returning NULL. However, when I run in the client:
I don't get NULL, I get an integer value, which is what the column is.
|
| Comments |
| Comment by Elena Stepanova [ 2018-08-03 ] | |
|
UNIX_TIMESTAMP works as expected, you're getting NULL because of
which, indeed, is quite obviously incorrect. How you got May 32nd on the slave – that's the question. Could you please quote the whole binlog event from the binary log (including variables which are set for the event)? | |
| Comment by Brian Kolaci [ 2018-08-03 ] | |
|
Wow, I looked at this for quite awhile and didn't notice May 32nd. This software has been running in production for over 10 years, but didn't fail until Mariadb 10.0. Looking at this I can definitely say its a bug in the policyd code, not the database code. Just amazing that its been working so long. I'll close this and open a bug with them to fix the SQL code. It should be an easy fix. They should be using date math and not: UNIX_TIMESTAMP(NOW()+2000000) (which appears to try to add 2 days). | |
| Comment by Brian Kolaci [ 2018-08-03 ] | |
|
Not sure how to close this. Thank you so much for pointing out something so obvious... |