Since my latest server crash / death, I'm planning on having a better plan for doomsday disaster scenarios. It took me 24h to be able to get a new server up and running to replace the dead one... Sucks because it was the webserver + databases. I'm feeling so happy because at least the webserver sites + db backups were being done using ISPCONFIG but into an outside drive... at least I had that So... I'm thinking on having an extra standby server for this emergencies. But have a few questions and would appreciate some recommendations. 1. With ISPCONFIG I can create a new server and make a mirror of another server. A. Ok, this would imply having a mirror server per server... kinda costly? Would it be possible to have it mirroring all servers including the panel, a full standby server capable of replacing any individual server? B. Another issue would be... an easy way to activate this mirror into a full fledged server in case of one of the servers crashed. With just the specific function of the dead server (email, panel, webserver, etc). The panel server (master database) points to different server id's and functions... the only way I could fix this was to mess with the master database and point all references from the dead server_id to the new server_id. Is there a better way to do this? 2. Even having 1 mirror server per server, is there a fast way to activate the mirror in the master and make the master understand that all functions related to the dead server_id now belong to the new server_id without hacking the master db? 3. But maybe I'm going all wrong about this! According to you guys, what would be the best and fasted way to recover a dead multi-server?
Syncing a multiserver setup to a single server will most likely require changes in the code. The easiest way is to use virtualisation e.g. with openvz. On each hardwrae node, run just a minimal server installation with openvz. Then create a openvz VM instance and install the ispconfig server into that vm. Then do regular backups e.g. with the vzdump command and store them on your backup server. In case of a crash, all you have to do is to setup a minimal OS with openvz kernel again and then use vzrestore to restore the full openvz backup from your backup server. Such a restore should not take more then 30 minutes. If your backup server runs a openvz kernel as well, then you can even start the openvz vm directly on the backup server, so it will take you just a minute to get online again.
Yeapsss, that's definitely a different approach, sounds like a plan. Do you also run the websites + databases on a openvz? Asking because I would not like the websites to gain page load times and times for first byte. I think I will give it a go with a secondary dns server.... I found this tutorial...might help someone... create a opnvz container from a physical machine: http://openvz.org/Physical_to_container If anyone else knows of a good tutorial to move a physical machine into a openvz container... please let me know.
I run everything in openvz. openvz is a kernel level virtualisation similar to a jail, so you dont have the huge overhead that other virtulisation solutions provide. Openvz is espaecially made for hosting servers.
OpenVZ Sucks Unfortunately openvz debian packages have been removed from the repositories since ubuntu 10.04. Code: * OpenVZ is supported on Ubuntu only for the 8.04 version. Code: Note that KVM is the main virtualization technology supported in Ubuntu. Main reason is the new kernel 3.x support for all types of virtual containers that openvz does not want to use. https://help.ubuntu.com/community/OpenVZ This solution totally sucks for ubuntu users.
Already added to bugtraker in version 3.0.4.6 http://bugtracker.ispconfig.org/index.php?do=details&task_id=2409
Works fine on debian stable and debian testing and thats the lnux distribution were ISPConfig supports openvz for.
