Regards managers,
Our planned scheme worked perfectly and we have now two identically
configured cluster hosts. Hopefully we avoid trouble regarding NetWorker
in the future. Digital, this is for you: Hopefully TruCluster in the
near future also can share tapedrives so that DECnsr is not single point
of failure but can be switched between cluster hosts as is the case of
disks. I also assume this require some changes to the DECnsr software
...
Thanks to Tom Blinn <tpb_at_zk3.dec.com> for pointing out that it is VERY
important to remember creating a symlink in usr pointing to /var (ie
lrwxrwxrwx 1 root system 6 Jun 25 13:16 /usr/var ->
../var)
in place of the old /usr/var directory.
The original posting:
>
> An ASE cluster is being backed up by DECnsr. The two cluster hosts are
> configured differently regarding /usr and /var. The first cluster host
> have separate /usr and /var, the second have a large /usr containing
> both /usr and /var. And now to the problem:
>
> When relocating services from the first cluster host to the second and
> vice versa, the mountpoints become different on the second because /var
> is symlinked to /usr/var and thus the mountpoints become
> /usr/var/opt/ASExxx/mnt... instead of /var/opt/ASExxx/mnt... which is
> the case on the first clusterhost.
> NetWorker (DECnsr) notice this and the indexes have different entries
> depending on which clusterhost the service is registered to. An example:
> If a service (let's call it service1) is relocated from clusterhost1
> (separate /var and /usr) to clusterhost2 then the entry in the indexfile
> will contain both /var/opt/ASExxx/ase/mnt/service1 and
> /usr/var/opt/ASExxx/ase/mnt/service1 thus doubling the space needed for
> the indexes and complicating recovery procedures. We have also seen
> problems backing up cluster services because of this. We have considered
> reconfiguring clusterhost2 as one way of solving the problem:
> What is the most convenient way of reconfiguring clusterhost2 ? We have
> considered just making another fileset (AdvFS) under the usr domain,
> copy the data from /usr/var to /var (in single user mode of course),
> rename the usr domain according to clusterhost1, removing the /usr/var
> data and reboot. (Of course we also have to update /etc/fstab). Are
> there any pitfalls I have to be aware of regarding ASE (hardcoded
> references to /usr/var etc.) ? Other gotchas ?? Is our planned scheme
> feasible ?
>
> The cluster hosts run 3.2G and ASE 1.3.
--
******************************************************************
* Knut Hellebų | DAMN GOOD COFFEE !! *
* Norsk Hydro a.s | (and hot too) *
* Phone: +47 55 996870, Fax: +47 55 996342 | *
* Cellular Phone: +47 93092402 | *
* E-mail: Knut.Hellebo_at_nho.hydro.com | Dale Cooper, FBI *
******************************************************************
Received on Wed Jun 25 1997 - 15:20:26 NZST