Advice about migration

Discussion in 'Installation/Configuration' started by electronico_nc, Sep 19, 2018.

  1. Well,
    Thanks to Florian, I have now a working standalone ISPConfig server.
    The aim is to have an HA config that is still available and administrative-able if one server fails.
    Idea would be to migrate the old server (using migration tool) to a new mirrored cluster (where IP Failover is activated, so IP can be switched from/to main/mirror).
    But (from Till) I need a master cluster for databases, so that if main server goes down, another master will stay connected to mirror and ISPConfig stays active on mirror (modifications are applied).
    Before going further :
    Do I have well understood ?
    Thanks
     
  2. florian030

    florian030 ISPConfig Developer ISPConfig Developer

    you can run two masters and make changes on server 1 or server 2. i don't think, that you really need this (and/or load-balancing). i prefer one server with full backups instead of a mirror-server. if (!) a server goes down, i can restore the backup.
     
  3. Thanks Florian,

    I only manage the technical side on servers (not their content).

    Unfortunately, some resellers have left 8 years old Drupal websites online without any updates... That must happen : sites have been hacked, they gained SSH access and launched DNS attacks.
    ISP has then rebooted server in FTP recue mode to prevent their security.
    It was the Master of the mirror.
    Thanks to the mirror setup, nearly everything was running on the slave.

    I learn each day and try to avoid situations that have been problematic.
    I know that the mirror setup has great pros (just discovered the cons about a live database master).
    If you restart a server after a kernel upgrade and it's time for an HDD fsck, you're offline for a while (it takes ages on 4T HDDs).
    Same thing if ISP has to change an HDD for maintenance.
    It's not possible to restore any backup under these circonstances.
    Services have to be quickly accessible, I would say maximum 3 minutes after any failure.

    I'm really ready to hear things that might help to reach the nearly ideal HA setup.

    Thanks for reading and suggestions.

    PS : I'm not ready to put sites on an obscur cloud instance that I won't have control on.
    Important services are : DNS, sites, email
     

Share This Page