Note: "permalinks" may not be as permanent as we would like,
direct links of old sources may well be a few messages off.
Daniel Khan schrieb:
> It's a 3 node cluster for a webhosting environment:
[...]
> node1 and node2 mount there home directories over NFS from node3.
> additionally node1 and node2 are keeping a mirror of their (NFS) data on
> node on their local disk.
1) DRBD is a _two_ node cluster
2) You cannot mount the DRBD-device in Secondary state
What's possible in your case:
A) split /home
node3 node1 node2
home1 Primary Secondary unconfigured
home2 Primary unconfigured Secondary
B) "chained"
node3 node1 node2
home Primary Secondary unconfigured
after failover you make
node3 node1 node2
home unconfigured Primary Secondary
Making node2 Secondary in this case will need a full sync. Depending on
the size of your device/partition and on your hardware, a full sync will
need more or less time to finish.
> If the setup above is too risky I'll use rsync based replication for now.
The advantage of sync utilities like rsync/unison on file or content
level is, that there is no restriction in direction or mounting.
The disadvantage is, that rsync cannot sync continiously. If you start
it e.g. by cron each hour, data changes of the last hour can be lost. In
case of read only webpages, your users maybe can live with this risk.
For web-forums, wikis, web-shops etc., where updates happen often, I
would prefer DRBD.
Helmut Wollmersdorfer