[MDEV-10398] segfault in mariadb 5.5.50 with galera 25.3.16 Created: 2016-07-19 Updated: 2020-10-20 Resolved: 2020-10-20 |
|
| Status: | Closed |
| Project: | MariaDB Server |
| Component/s: | Galera |
| Affects Version/s: | 5.5.50-galera |
| Fix Version/s: | N/A |
| Type: | Bug | Priority: | Major |
| Reporter: | Evan Jardine-Skinner | Assignee: | Unassigned |
| Resolution: | Incomplete | Votes: | 0 |
| Labels: | None | ||
| Environment: |
A 3 node cluster using Ubuntu 14.04 with mariadb 5.5.50 and galera 25.3.16. |
||
| Attachments: |
|
| Description |
|
segfault during normal operation of mariadb. Right at the bottom of syslog. I had hit this before so I had turned on the general query log. I've attached the general query log, syslog (with most of the irrelevant stuff, like CRON etc filtered out) and the last binary log from the node that crashed. Let me know if you need more files from the machine. It will liekly be reinstalled in a few days. |
| Comments |
| Comment by Elena Stepanova [ 2016-07-19 ] | ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
|
Ev, hi, if you can afford an experiment (at the risk of getting the server crashed again), could you please check whether the following sequence executed from a new MySQL client session causes the crash?
In any case, please attach your cnf file(s) and paste the output of
If you can provide the datadump of the users table (or .frm/.MYI/.MYD files if it's MyISAM), it would be even better. You can upload them to ftp.askmonty.org/private. Thanks. | ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| Comment by Evan Jardine-Skinner [ 2016-07-19 ] | ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
|
I will try those commands asap. Here is a dump of the mysql database in the meantime. And also jsut he suer table which is what you actually wanted | ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| Comment by Evan Jardine-Skinner [ 2016-07-19 ] | ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
|
Output of:
is
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| Comment by Elena Stepanova [ 2016-07-19 ] | ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
|
Ev, Thanks, but the dump is actually from a different table. The query is from healthcheckcat_5.users, not from mysql.user. | ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| Comment by Evan Jardine-Skinner [ 2016-07-19 ] | ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
|
Oops. Many apologies I was in a hurry then before getting to a meeting. Am uploading a tgz of the /etc/mysql directory and also the dump of the correct table | ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| Comment by Evan Jardine-Skinner [ 2016-07-19 ] | ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
|
Running the commands in your first comment does not cause mysql to crash. I see the following:
And afterwards I still have Primary status for the node | ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| Comment by Elena Stepanova [ 2016-07-19 ] | ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
|
That's rather bad news... If it's not readily reproducible on your side, it's going to be difficult to reproduce on ours. Do you happen to know where connections which run this query actually come from? Given that the connection that crashed executed ~70 queries in 2 seconds, it's probably not a client, but some connector, do you know which it is? Please also paste the output of
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| Comment by Evan Jardine-Skinner [ 2016-07-21 ] | ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
|
Here's the dpkg output
This output has reminded me that these servers are actually Ubuntu 14.04 and not Debian 7 as I originally mentioned. Sorry. | ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| Comment by Evan Jardine-Skinner [ 2016-07-21 ] | ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
|
The code that is running the queries comes from three places: 1) Aegir (site management for Drupal): creating and deleting drupal web sites. Creating the site involves restoring a database (I think via the mysql cmd line client) then changing some values in the database (via mysql cmd line) and running a bunch of php to do other configuration stuff to the DB. It then immediately kicks off the cron tasks referenced in 2) 2) Drupal 7 websites (no users are navigating around the GUI as this is a test system): running cron tasks (in PHP), which do stuff like build up caches of data, insert data into the drupal db etc. 3) Some system scripts that do various checks and use DB tables to create lock rows so that conflicting tasks do not happen at once. These all us the mysql cmd line client. | ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| Comment by Elena Stepanova [ 2016-08-28 ] | ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
|
Stack trace from the syslog (to make it searchable in JIRA by whatever little it has):
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| Comment by Elena Stepanova [ 2020-10-20 ] | ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
|
We have never been able to reproduce it, the stack trace is not distinctive enough to match it with other open or fixed bugs, and 5.5-galera is long gone now; so I'm closing it as incomplete. The issue can be re-opened if it gets fresh information related to active versions. |