[Veritas-bu] Backups slow to a crawl
2005-03-24 18:16:31
Jeff,
On Thu, 24 Mar 2005, Jeff McCombs wrote:
> Ok. I lied. Removing multiplexing did not fix the problem.
>
> It's strange, I _know_ my network is clean, I know my backup policies should
> be fine..
>
> I'm still concerned about the busy percentage of rmt/1 v.s. Rmt/0.
[...]
> In trying to troubleshoot, I watched the system's I/O performance using
> 'iostat' and noticed that /dev/rmt/1, the 2nd drive in our library (Overland
> Neo 2000) appears to be having some problems in sending data to tape. I
> noticed that the %-busy on the drive shoots up to 100% as kw/s (kbytes
> written/sec) drops drastically down into the 200-300 range.
>
> /dev/rmt/0 has no problems during the same time period. %-busy sits anywhere
> from 2 to 30%, and kw/s is in the 1.2 to 2.5 range.
>
> The only correlation I can find with systems that are failing backups with a
> 196 status are systems that were queued to rmt/1. Systems queued to rmt/0
> backup fine, and usually these systems backups complete in 15 minutes or so.
[...]
I know others have suggested replacing the drive. But first - have you
tried cleaning it? (eg. force it with tpclean.) I've observed extremely
similar symptoms, and cleaning has been known to sort it out totally. Now
if I could just work out why the cleaning didn't happen automatically ...
HTH,
Richard
|
|
|