[MDEV-11247] Crash: mysqld: Out of memory (Needed 40 bytes) Created: 2016-11-07 Updated: 2016-12-27 Resolved: 2016-12-27 |
|
| Status: | Closed |
| Project: | MariaDB Server |
| Component/s: | Server |
| Affects Version/s: | 10.1 |
| Fix Version/s: | N/A |
| Type: | Bug | Priority: | Minor |
| Reporter: | Martin | Assignee: | Unassigned |
| Resolution: | Cannot Reproduce | Votes: | 0 |
| Labels: | None | ||
| Description |
|
My journalctl for the whole day is included in the pastebin log. mariadb 10.1.18-1 Starting the service again worked and it is my first crash ever. The RPi3B isn't running many things so it is using only about 200MB RAM at the time(out of 1GB). |
| Comments |
| Comment by Elena Stepanova [ 2016-11-07 ] |
|
~800 MB for 10.1.18 is really not much, and I guess if it chokes at 40 bytes, it is quite possibly that one start failed while others squeezed into the available memory, although just barely. It's not clear from the description whether the configuration has been tuned for low memory consumption – mostly it means reducing buffers, especially less used ones, disabling unnecessary engines and plugins, etc. |
| Comment by Martin [ 2016-11-07 ] |
|
The myslqd process is running at some <80MB RAM usage. I believe I didn't change any default mariaDB configuration. |
| Comment by Elena Stepanova [ 2016-11-07 ] |
|
mysqld is incapable of running at 80MB, especially with default configuration. InnoDB buffer pool size alone is 128M, and there are some other equally big buffers. The more realistic size is ~750-800MB. So, I guess something is wrong with the measurements. You need to tune configuration if you want it to run reliably on your system. It's totally doable, it just requires some non-defaults, and which ones to set depends on your workflow. I suppose you can search for some tips in the internet, and/or ask on the mailing lists. |
| Comment by Martin [ 2016-11-07 ] |
|
http://i.imgur.com/EgPs3UR.png Seems to be using only about 80MB of RAM to me. I'm only using MariaDB for mediawiki, phpmyadmin, mantisBT and probably one or two more, and I could count concurrent users on all of those combined on a single hand. I'll try looking up what settings I should try changing then. |
| Comment by Vladislav Vaintroub [ 2016-11-07 ] |
|
It looks like the process is using 1.5GB virtual memory from the picture. |
| Comment by Elena Stepanova [ 2016-12-09 ] |
|
C0rn3j, have you been able to configure the server to use less memory? As wlad said above, from the picture that you provided it does look like the process is using ~1.5G memory, so, given our hardware limitations, out of memory is justified. |
| Comment by Martin [ 2016-12-09 ] |
|
This issue stopped happening on mariadb 10.1.19. Either some bug was fixed or the maintainers changed some config ^^ |
| Comment by Elena Stepanova [ 2016-12-27 ] |
|
Since the margins were so tight, I suppose it's quite possible that in the latest release some scenario important for this instance started working different so that it consumed a little bit less memory, enough for the instance to stay within limits. I'll close the bug for now, but you can re-open it later by commenting if it starts happening again. |