On Tue, Jul 09, 2019 at 12:55:18PM +0200, Jan Zerebecki wrote:
On 09/07/2019 10.35, Daniel P. Berrangé wrote:
If virtlogd is active, we do *not* want logrotate doing anything at all.
Are you saying that to fix the bug at hand, I should be required to first implement 2 features in virtlogd and replace the current logrotate config (none of which caused the bug)?
I'm simply saying that trying to use logrotate at the same time as virtlogd is not a desirable way to solve the problem. There are a variety of other options that could be explored
Trying to get sensible interaction between two separate log rotation apps is adding too much complexity.
It is not added. It is already there. Multiple distros I checked ship this as the default config. Sorry, if that wasn't your intention.
Shipping the configs is not a problem as long as make the logrotate config a no-op when virtlogd is active, which was the intention and is fixed in: https://www.redhat.com/archives/libvir-list/2019-July/msg00313.html
Further, any purging of log files needs to take in to account what guests actually exist in libvirt. The proposed change has no such checks. It must only purge log files for guests which are neither running, nor have any persistent config on disk, and where the log file is older than "N days" for some configurable "N".
Not if time based rotation is already desired, then there is no need to know if the guest still exists. If its not desired, then this is as complicated as you say, but the existing logrotate config already doesn't fulfill that.
It will no longer be time based with the above fix. Regards, Daniel -- |: https://berrange.com -o- https://www.flickr.com/photos/dberrange :| |: https://libvirt.org -o- https://fstop138.berrange.com :| |: https://entangle-photo.org -o- https://www.instagram.com/dberrange :|