Backup/restore behavior in a multi-server replicated setup

Discussion in 'Installation/Configuration' started by StackFault, Oct 2, 2015.

  1. StackFault

    StackFault New Member


    Some background info to start:
    I have a multi-server ISPConfig setup running. The setup is made of 3 web/db servers. (1 master and 2 slaves)
    Database is a Galera Multi-master cluster and the file system is on GlusterFS.
    Filesystem and database is synchronized between all servers in real time.

    My problem:
    When I configure backups for a site, every servers (master AND slaves) takes backups while only the master should, in my opinion. Same thing on the restore, all servers tries to restore while only the master should. This results in database lock and conflicts since everyone tries to write on the same records at the same time, same thing for the files.

    My assumptions:
    Since we need a replicated db/fs for the multi-server setup to work, the backup/restore process should take place only on the master server. Restored data will make it's way to the slaves by the db/fs replication.

    My questions:
    I have looked around and didn't found a way to fix this in an elegant manner. Is this issue addressed in the coming release or there is a configuration I am not aware of?

    I already know I could change the code to prevent slaves from restoring or taking backups but IMHO this should be fixed at the source instead.

  2. florian030

    florian030 ISPConfig Developer ISPConfig Developer

    With the current version you mst remove the backup-code from cron_daily.php (or mayber just disable on the slave-servers and remove /usr/local/ispconfig/server/plugins-enabled/ from the slaves. The upcoming-version has a class for each cronjob so you can disable single jobs like web-backups.

Share This Page