Re: questions about hot backups/archived logs

On Thu, 10 Apr 2003 17:37:13 +0000, NetComrade wrote:

> We are currently not running in archive log mode.
> We have 13 databases running on 3 hosts (e4500) attached to 4 disk
> arrays (a5200), all data is striped and mirrored including redo logs
> (kind of like SAME).
>
> I don't want to put archived logs on the same disk arrays (extra io
> contention, file systems will get fragmented, it's a vxfs), so we are
> thinking of possibly putting them on NFS (we have an NFS cluster)
> However, I have certain concerns about NFS (it being unreliable, etc),
> so I'd like to know if anybody else uses NFS for archived logs.
>
> Additionally, we want to put backups on nfs, however, that will bring
> in extra network contetion, especially if we won't be able to compress
> datafiles on the fly.
>
> Our databases range in size from 5 to 80Gig (300Gig overall), and
> generate up to 100G of redo per day (that includes redo wastage)
>
> I am exploring the possibility of freeing up one disk array and using
> it exclusively for archived logs, but that might prove difficult,
> additionally, i would still need a place to backup 300G of datafiles.
>
> Any advice appreciated.
> .......
> We use Oracle 8.1.7.4 on Solaris 2.7 boxes
> remove NSPAM to email

I think nfs is ok as long as you ensure that nfs does actually write the data through before returning ok to the clients. Degrades performance, but its a safe bet.
Settig up a dedicated network connection to your backup/nfs systems might be a good idea (Gbit line?). In that case you do not disturb normal network traffic.

What about decoupling oracle archiving from nfs access. Setting up a offline redo log stage area where you hold only a fraction of your offline redo logs. Oracle does write the offline redo logs to a local disk. A cron job can take care of getting the offline logs ( e.g. all logs older than the current 10 ) to the nfs cluster (again nfs configured to play safe).  

-- 
Juergen Lueters

jlueters_at_intranet-engineering.de