tapes prematurely full

[ElHedi]

Dear All,
I have the following configuration :
networker server 7.2 on windows 2003SP2
networker dedicted storage nodes : HPUX
networker clients windows and linux backing up on the server .
I have prematurely full tapes especially on windows and linux and not HPUX.
searching on your forum I found some posts and tried the following :
-installed the latest IBM certified drivers on windows
- disabled CDI
-cleaned the drives
-put news tapes
the problem still exists so I am thinking about block-size problem can somebody help me how to troubleshoot that?
thanks

 

[bombcan]

Hi,

In my experience I found there were hardware issues. Check your event viewer for any hardware problems on the windows backup server.

 

[a060463xyz]

Which library do you use ?
A blocksize mismatch usually leads to tape positioning problems. Networker will mark Tapes as full if there are SCSI problems.
If you are using FC check your zoning.
Disable TUR on Windows

 

[ElHedi]

I am using Adic scalar i2000
I ran scaner -vvv on several volumes I found 65k
on devices it is set to handler default
I am going to disable TUR.
I checcked Zoning it is ok.

 

[605]

The block size is also written in the label. Consequently, changing the block size without relabeling the media does not really make sense.

NW marks the media full if it can not recover from a hard write error. This has nothing to do with the block size ... as long as the driver allows to use the block size which is found in the media.

If you for example label a tape on HPUX it might be that you will not be able to write it on Windows because Windows usually must be 'patched' to allow larger block sizes than 64kB. Please see the "Performance Tuning Guide" for details.

However, i don't think that such condition will cause a write problem resulting in a prematurely filled tape if you try to write a tape with a larger block size than 64kB on an 'unpatched' Windows SNode.

If each tapes can be used without problems in their 'own environment', i.e if you can successfully label a tape on Windows and fill it nicely on a Windows node, then i do not expect a general hardwarde problem - the problem is most likely due to interchangeability.

 

[ElHedi]

Thank you 605
I disabled TUR
cleaned drives.
things are still the same.
Now some tapes, after being written to, become suddently unusable so I tried the following on the same drive
nsrjb -l -n -f <device> <volume> : ok
mt -f <device> status : ok
nsrmm -pv <voulme> : failure

this means that lebels have been overritten ?

Other tapes succeed in those tests but they are getting full also.
Can I say that it is sure a hardware issue as Bombcan said it?
are there other things to check before changing drives?
May scalar firmware be the cause in some way?
the tapes are new so I think the issue is not with them
thanks

 

[605]

It looks like a HW issue. If you want to make sure, use another program (tar?) or NW's tapeexercise command which might help you.

However, although it is a good idea to install the latetst F/W update, i do not think that the one for the jukebox matters. Of course the one for the drives is important.

 

[ElHedi]

I have Emulex HBA on server.
I have confih=gured persistent binding on HBanyware side,
have I to configure persistent binding on OS side too?
thanks

 

[605]

Persistent bindings will especially help for systems which will auto-reconfigure after a reboot. In such cases, the OS might assign new device names which will screw up an existing configuration.

Persistent bindings is a tool to permanently bind the same device name to a certain hardware address.

Unfortunately, i do not know what to do on the OS side. To the best of my knowledge, using one method is enough. It seems that the feature which was originally designed at the driver level is now also available at the OS level.

Now - with respect to your problem - i do not think that the two issues are related. One the device is accessible, this is it .. until the hardware will be auto-reconfigured after a reboot.

 

[a060463xyz]

do you have a path mixup ?
run inquire on every storagenode and notice the serialnumbers and WWNs of the tape drives.
Check if the drive order is changed on a storage node.

Do you have FC or SCSI-drives in your scalar ?
With scsi check the termination.

 

[SDack]

Are you sharing tape access with drive acess in the same HBA?

 

[SDack]

Another question, what type of LTO are you using?...

 

[ElHedi]

Hi SDack,
drive access are not on teh same HBA as disks.
We have LTO2 drives.
We had installed the IBM drivers in non exclusive mode since that time problems began.
We reinstalled the same drivers in exclusive mode. The problems seem to be stopped since 10 days.
This is strange because EMC recommands non exclusive installation.

 

[d3h614]

we had this problem -- tape device drivers is pretty much always the problem.