Cyrus Proxy Maint.
When the mupdate databases on the non-master nodes become corrupt or out-of-sync you'll need to bring the proxy down and take steps to re-sync the database from the master node. If you need to reconstruct the master mupdate database it's important to understand that the procedure for that is different.
Procedure 1. Resync and restart proxy
-
First log into the affected proxy node. Once there cd into /var/cyrus/imap as this will be the primary directory where work will be done.
-
Next identify the pid of the master process.
ps -fade|grep master|grep -v grep
If the box is so loaded that the ps takes too long to return, cat /var/run/cyrus-master.pid should return what you need.
-
Next, kill off the master process
kill <pid>
![[Note]](../../../graphics/admon/note.png)
This might take a while This step might take a little time due to the number of child processes that the master process may have.
-
Remove the old db files and the core file that was dumped when master was killed.
![[Warning]](../../../graphics/admon/warning.png)
Check your PWD Make sure you're in /var/cyrus/imap when you run the following rm command
rm -rf db/* db.backup* proc/* socket/* mailboxes.db tls_sessions.db core
-
Next you'll need to comment everything except the mupdate service from the service section of /etc/cyrus.conf so that when the master process is started only the mupdate subservice is started.
-
Edit /etc/imapd.conf to uncomment the following line:
#skiplist_unsafe: 1
-
Restart master:
/usr/local/bin/master -l 512 -d
-
Watch the imapd log for the completion of the database syncronization.
tail -f /var/log/imapd.log
-
When the database syncronization is complete kill off the master process again.
-
Recomment skiplist_unsafe: 1 in /etc/imapd.conf
-
Uncomment the services in /etc/cyrus.conf that you commented a couple steps ago.
-
Bring the master process back up.
/usr/local/bin/master -l 512 -d
This might take a little time and if you check /var/log/imapd.log for progress you'll see the mupdate database being brought back online which takes about 30 seconds. Once it's back keep watching the log for proxy and other activity.

