log rotation support

[gryphius]

When using file logging it would be nice to have support for logrotation, based on size or time. Currently we use logrotate for this - but it seems like cascaded requires a full restart after the logrotation to start writing the new file. Maybe there could be a command like cascade logs reopen or something like that to recreate the filehandles without a full daemon restart?

Or maybe cascade could do the rotation itself. It seems like the used trace module does support rolling logs.

[ximon18]

Hi @gryphius,

While I appreciate that this would to some extent be convenient, isn't this better handled by a dedicated tool that has more features, is proven, and doesn't expand the amount of code in or number of dependencies used by Cascade?

Ximon

[gryphius]

Hi @ximon18
I fully agree, but I realize I may have worded this badly: We are currently using logrotate:

/var/log/cascade/cascade.log {
    daily
    rotate 7
    compress
    missingok
    notifempty
    create 0755 cascade cascade
    postrotate
        systemctl restart cascaded || true
    endscript
}

instead of systemctl restart cascaded I would prefer to have a command that tells cascade to re-open /var/log/cascade/cascade.log without the need for a full restart, similar to let's say BIND's rndc reconfig. Restarting cascaded takes a long time with large zones so I'd like to avoid that if possible.

[ximon18]

Thanks @gryphius, my bad, I read the title on my phone and stopped reading rather than open the full issue.

Indeed a full restart should be avoided unless absolutely necessary.

Would the copytruncate option not also solve your problem?

Ximon

[ximon18]

Perhaps the warning about data loss with copytruncate is reason enough to want a way to tell Cascade to close and re-open its log file. I wonder how we handle this for our other products such as Routinator, Krill and RTRTR.