NetWorker says device is writing, but it is not 1

[asimpson]

Yesterday I started a test backup of a new client. I canceled the backup after about 30 minutes of seeing that the tape drive had been stuck at "writing at 4161 KB/s, 381 MB, 2 sessions". After I canceled the backup, the tape drive still showed "writing at 4161 KB/s, 381 MB, 2 sessions". There were no stray nsr processes for the client, however, the client still showed up in the sessions window, once for C:\ and once for SYSTEM FILES:\. At this point a day later, the tape drive still shows "writing at 4161 KB/s, 381 MB, 2 sessions" and absolutely nothing belonging to the same Pool as the new client I was testing will backup. I know that I can clear this right up by restarting NetWorker, but there are too many other backups running at this time. I cannot unmount the tape through NetWorker because it says "device is busy". I cannot run nsrjb -H to reset the jukebox hardware because the first tape drive actually is writing so the command won't go any further. Is there any command at all that anybody can give me to run to clear up this confusion that NetWorker thinks the tape drive is writing when it obviously is not?

 

[605]

You can actually start the nsrmmd process ... if you exactly
know which one is controlling that specific device.
NetWorker will try to restart that daemon after 2 minutes
(default).

Please also keep in mind that netWorker will not delete
all messages right away.

 

[asimpson]

I don't know exactly which nsrmmd process is controlling the device. Is there a way I can find out?
Also, it's not a message that I'm concerned with, it is that the client is still listed in the Sessions window even after the backup was canceled, and the tape drive still says 'writing 4161 KB/s, 381 MB, 2 sessions' and has not changed for 24 hours. I can actually dismount the tape through ACSLS and NetWorker will still show the drive as writing 4161 KB/s, 381 MB, 2 sessions.

 

[605]

All nsrmmds on UNIX/Linux have sequence numbers which you
should see at least with NW 6.x and upward - whenever a new
nsrmmd is started, it will get the next higher number assigned.

Usually, NW will just start the devices in the alphabetical
order, so it should not be a problem to build up the
relationship. Some more information are also available from
the file /nsr/logs/daemon.log.

I have never worked with ACSLS but know that the robotic is
controlled via this server. This is why you do not need
NW to unload the media. So i assume if you execute an unload
for the drive, it will just work. However, if the nsrmmd
process hangs, i do not expect a valid status message - the
only thing you can hope is that NW will cut the save process
after nsrexec has reached the timeout value (.. -T 1800),
which is nothing else but 30 minutes (default), configured
for the group.

 

[Barks]

If your backup server is UNIX, try the fuser command.
Ex. fuser /dev/rmt/0mnb
This will list the process associated with the tape drive which you can then kill. The nsrmmd process will restart and then you should be able to unmount the drive.
FYI, there is another note in this forum that discusses this issue as well.

 

[mordrin]

If you use Networker in a Windows environment, I can recommend the freeware tool Process Explorer from Sysinternals (

http://www.sysinternals.com/ntw2k/freeware/procexp.shtml).

It can show open handles by process, so if you do a search for the tape device in question, you can find the matching nsrmmd.

 

[Rickardo]

asimpson!

I have a similar problem.

Backups hangs. The only way to unhang it is to run the cmd "inquire -c". Does this work for you aswell?

Im intrested in what kind of hardware your running.

My setup is a Sun server running solaris.
2 HBA's (Emulex 9802)

connected to:
An ADIC Scalar i2000 with IBM LTO2 drives

The funny thing in my case is that the drives to one HBA card is hanging. The other two still works fine.

Anone have any sugestions what this could be caused by?

Tried a new HBA card, upgrading firmware on the HBA cards.

// Rickard

 

[605]

Forget NW for a while and just investigate the devices. Do not proceed before your 'normal' OS backups (tar etc.) will work successfully.

 

[HelpNow]

I have seen this situations in windows client of networker server where client memory or server load is so high for any number of reasons that the server failes to backup this client for long time until the services are restarted on the server side the sessions are not going away, i have not tried restarting client side, but that might work.
Issue to identify the nsrmmd process controling the device on server side still remains.