Note: "permalinks" may not be as permanent as we would like,
direct links of old sources may well be a few messages off.
Dominique Chabord wrote: > For some studies about disaster tolerance, I'm planning to implement > VSERVERs replication with DRBD, I mean replicating the full vserver > directory. Is it what you are using ? I am not using vserver in production. I only tried out the setup and configuration. I also want to try XEN. After this and some testing, I will decide, if I consilidate a third physical server into my heartbeat+DRBD cluster. And yes, the plan is, to have the full vserver directory on a DRBD resource, which allows automatic failover of a vserver guest by heartbeat. More flexible would be, to have each vserver guest on its own DRBD device. > We use Debian Sarge kernel 2.4.29 patched, three hardware servers and > ten vservers. Our objective is to guaranty all ten vservers are always > up once and there is always a replicated copy of them among the three > vservers. We now have all the pieces, but never integrated the whole > picture. > my questions: > How reliable vserver restart on replicated volume has proven to be ? I restarted several times, but I didn't prove the details. From DRBD I would not expect problems. The problems I had, were typical vserver newbie problems. The only problems are the limitations of vserver. > Any hints before we start digging any further ? I would change to a recent kernel like 2.6.11.11 and the latest vserver. Additionally I would take XEN into account, which seems to have only a little bit more CPU overhead compared to vserver. The decision between XEN and vserver depends on details, and if your configuration can have benefits of some details. In my case it can be an advantage, that the most important developer 'Bertl' of vserver lives near the same town, and I know him face-to-face. Helmut Wollmersdorfer