Details
-
Bug
-
Status: Done
-
High
-
Resolution: Done
-
2.14.0
-
5
-
Yes
-
Yes
-
Yes
-
[obsolete] C/S Core
Description
As a result of issues such as PMM-7464 the log files for NGINX can grow out of control, e.g.
$ du -sh /srv/logs/nginx.error.log 13G /srv/logs/nginx.error.log
The log managed by supervisord is already controlled, but the ones left to NGINX are not.
Suggested fix:
- add log rotation to allow log control
Implementation:
- add log rotation
- delegate log rotation to exclusively to supervisord
- redirect all nginx logs to /stdout and /stderr vs to log files
How to test:
- there are automated tests, which should be green:
https://github.com/Percona-Lab/pmm-api-tests
https://github.com/percona/pmm-managed - get inside of the PMM docker container and verify that the `/srv/logs` file list is limited to the following:
- a fresh install wouldn't have the files, deleted by this PR
- a PMM Server upgraded from an older version to this one would still have those files, but this is impossible to test
Example of a PMM Server latest version:
Attachments
Issue Links
- blocks
-
PMM-8307 Default config limits for allowed connections prevents monitoring large numbers (500+) of DB servers
-
- Done
-
- duplicates
-
PMM-5845 Logrotate issues with Nginx/Supervisord logs
-
- Done
-
-
PMM-5385 Manage all PMM Server logs with supervisord
-
- Done
-
- is duplicated by
-
PMM-5292 Reduce amount and size of ClickHouse logs
-
- Done
-
- relates to
-
PMM-7464 NGINX misconfiguration leads to log storm in push mode
-
- Done
-
(2 mentioned in)