Infrastructure/ansible
1
0
Fork 0
Code Issues Pull requests 6 Projects Releases Packages Wiki Activity Actions

copr-be: we shouldn't use both cronolog and logrotate

Browse source 
These implications constraint us:

    max-workers is used => cronolog needs to be used
    log counter enabled => we need logrotate

But:
    when logrotate is used => cronlog doesn't detect rotation

So the '4c4b63516d' is IMO explained.
Cronolog kept writing the log to the old descriptor, and the new
access.log link was just empty (before 4c4b63516d when we just
reloaded, not restarted).  So I hope now that there isn't any other
problem that would require us to restart Lighty regularly.

The **UGLY** work-around is to terminate the cronlog process.  As I've
found out, cronlog is automatically restarted by mod_accesslog - though
this is somewhat gray zone and we should long-term do one of those:

    - stop using logrotate && rework log counter
    - stop using max-workers && cronlog
    - stop using Lighty

Resolves: https://pagure.io/copr/copr/issue/2001
Resolves: https://pagure.io/fedora-infrastructure/issue/10391
This commit is contained in:
Pavel Raiskup 2021-12-03 17:03:03 +01:00
parent b005356b10
commit 19b8f82f5b
2 changed files with 16 additions and 6 deletions
Download patch file Download diff file Expand all files Collapse all files

7
roles/copr/backend/templates/lighttpd/lighttpd.conf
View file

@ -532,7 +532,12 @@ $HTTP["url"] =~ "^/archive/spacewalk/" {
index-file.names = ( "index.html" )
accesslog.filename = ""
}
server.max-worker = 6 # this may break many things
# This may break many things, per: https://redmine.lighttpd.net/projects/1/wiki/Server_max-workerDetails
# At least we have to use cronolog: https://redmine.lighttpd.net/projects/1/wiki/Docs_ModAccesslog
# But we need logrotate for logcounter, see /etc/logrotate.d/lighttpd, and
# logrotate doesn't work well with logcounter.
server.max-worker = 6
$HTTP["url"] !~ "^/archive/spacewalk($|/)" {
accesslog.filename = "|/usr/sbin/cronolog /var/log/lighttpd/access.log"

15
roles/copr/backend/templates/logrotate/lighttpd.j2
View file

@ -1,7 +1,12 @@
# We don't want to reload lighttpd twice (for each log file), so we use
# the sharedscripts option. This means that hitcounter is run only once too
# (which is good). In (likely impossible) situations when access.log is empty
# and error.log not, we would call hitcounter too on an empty file.
# We don't want to reload twice (for each log file), so we use the sharedscripts
# option. This means that hitcounter is run only once too (which is good). In
# (likely impossible) situations when access.log is empty and error.log not, we
# would call hitcounter too on an empty file.
#
# Note that lighttpd server runs with max-workers, and thus we pipe the access
# logs through cronolog to the access.log. It doesn't make sense to -HUP
# Lighty, because then it doesn't restart cronolog. We need to -HUP the
# cronolog directly (which is fortunately re-started by Lighty/mod_accesslog).
/var/log/lighttpd/*.log {
rotate 5
@ -18,6 +23,6 @@
{% endif %}
endscript
postrotate
/usr/bin/killall -HUP lighttpd &>/dev/null || :
/usr/bin/killall -HUP cronolog &>>/var/log/copr-backend/cronolog-restart.log || :
endscript
}