StoreEver Tape Storage
cancel
Showing results for 
Search instead for 
Did you mean: 

Re: Bad tape causes network slowdown

 
Highlighted
Occasional Contributor

Bad tape causes network slowdown

Configuration: We have an MSL5052 tape library with 4 SDLT drives. Our tapes are Maxell SDLTtape 1. We are running Dataprotector 5.5, just updated with latest patches on 9/24/07. The cell manager is a Proliant DL360 running Windows 2003 Server sp2. The tape library is connected to our EVA 5000 on the fiber fabric, and to some servers on fiber, and all servers on 100Mbps Ethernet.

Problem. About every month or two, usually while doing unattended full backups on the weekend, Dataprotector will report unrecoverable tape errors.

When this happens, rather than gracefully failing or resuming the backup with a new tape, the backup job stalls, and it takes down the whole ethernet network. No other jobs can run on the network when this happens, as it seems like the network crawls along about 300baud. When we come in on Monday morning, no one can log in.

Our solution is to reboot the MSL5052 (by powering down and back up). This clears the network problem, and allows others to login and work. We then use Dataprotector to abort the backup job and start over. The MSL does not have a remote console or ILO equivalent, so we have to be there in person to power down.

Most of the time the tape is bad - it fails to verify, but sometimes it will verify successfully. Nonetheless, there is no reason for a bad tape to affect the whole network!

* What can be causing this?
* What tests can I run to determine where the problem is? Is it in the MSL hardware, in Dataprotector software, in a NIC, or where?
* What can I do to prevent this from happening at the next bad tape?
1 REPLY 1
Highlighted
New Member

Re: Bad tape causes network slowdown

To troubleshoot, i would pull the network cable to isolate the issue.

I would also use a sniffer, to see what the NIC is trying to access.

In the mean time I would isolate the library on it's own network (vlan?). That way it wouldn't affect other users.