Note: "permalinks" may not be as permanent as we would like,
direct links of old sources may well be a few messages off.
> > > /dev/mapper/lvm-tmp on /tmp type reiserfs > > > (rw,noexec,nosuid,nodev,notail) /dev/mapper/lvm-usr on /usr type > > > reiserfs (rw,notail) > > > /dev/mapper/lvm-var on /var type reiserfs (rw,nodev,notail) > > > > When you want to have automatic failover with heartbeath, mirroring /usr > > won't work, at least not directly as /usr. > > But among the remaining 3 volumes listed above, I don't think I need > to mirror them because AFAIK, /tmp contains only temporary files; /usr > are binary and docs/manpages files; and /var for logs. Am I thinking > that right? Yes, I think so. > > > You need /var/lib/nfs on both systems to be equally, we did this by > > linking it to a directory on one of our mirrored partions. Theoretically > > there's also another (better) way with 2.6., but this is still not > > thoroughly tested. > > Well, I'm still configuring the master node and after I finish > configuring DRBD, I'll make a disk image copy for the secondary node > by using the dd utility. In my master node, I already installed This won't help for /var/lib/nfs, it is managed dynamically, so you really need this all the time to be consistant! > nfs-utils-1.0.6-2.1.101mdk.i586.rpm and related packages but I'm > thinking to use the nfs4-utils-1.0.6-2mdk.i586.rpm and it's related > packages. What do you think is best for NFS server package? I > haven't tried NFS 4 yet and I don't know if it's good enough for > production. Any advice? I don't know what mandrake is doing, but offifical nfs utils will have many updated to nfsv4 with 1.0.7 (to be released this week?). Anyway, NFSv4 is still a rather new protocol and it is not just a (slight) improvement as V3 to V2 was. Furthermore it is pretty much untested, not what one really wants to have in a high availibility environment... > To summarize, based on your advice, I'll just mirror only my > "/dev/mapper/lvm-home on /home type reiserfs (rw,nosuid,nodev,notail)" > volume. Yes, I think thats the only partion from your list that makes sense to be mirrored. Cheers, Bernd -- Bernd Schubert Physikalisch Chemisches Institut / Theoretische Chemie Universität Heidelberg INF 229 69120 Heidelberg e-mail: bernd.schubert at pci.uni-heidelberg.de