To CDI or not to CDI that is the question

Hi Jason,

I tried to do some research on this for you.  Usually if a drive goes to service mode there is a reason for it.  That reason should be investigated.  Many cases are because its due to bad tape device drivers on OS.  Most of the scenarios are documented in technical note “Configuring Tape Devices for EMC NetWorker” which is available on PowerLink.  A general rule of thumb is CDI disable should be used as a last resort.  For example, you are using bad drivers and you cannot upgrade/change them.

Hope this helps.

Allan

I've also not yet seen a case where it was recommended to have CDI enabled for virtual tape drives.

It's a great concept, but it seems to be a black art. I've even found different versions of OS' cause a problem, or a new version of NW with the same OS.

A great example of it not working was 7.4.? where we had an ACSLS library with DDS just between systems all running Solaris 10 and we couldn't get the config to work until we switched off CDI.

I've seen other circumstances where the only way I can get it to work is to switch it on.

basically, I always go with having it on, and then switch it off if I get problems which do not seem to be drive related.

Jason,

If you have issues with CDI turned on the proper path would be to call Support and have Support contact engineering to troubleshoot the actual issue. What is most liekly happening is that NetWorker is getting TapeAlerts from the device, virtual or not, in a redundant fashion. NetWorker will see this as a drive running erroneously and do what it deems as the corrective action to ensure consistency of your backups, and that is to disable what it thinks is a bad drive. If this is happening and we know that the drives are not bad, or the device driver in question is not a problem, we need to be able to troubleshoot the code of CDI to fix it is needed. By simply turning CDI we are not fixing the problem but rather hidding it. Sort of like unpluggin the light bulb on your check engine light in the car. You still have the underlying problem but it simply is hidden.

I would like to encourage you to keep CDI on as best you can and work with engineering via Support to resolve the actual issue. There are great benefits to CDI:

• Detection of device order changing via unique identification
  
  
• Enables jbedit, streamlines library config
  
• AS mentioned above- TapeAlert support...clear, decriptive information for faster error diagnostics and quicker time to resolution
  
• Support for persisitent SCSI reserve/release
  
• Future support, lke LTO-4 encryption, will rely on CDI
  

So, as you can see, there are good reasons to enable CDI and track any issues that may occur.

I welcome any and all feedback on this discussion as it wil only help us further understand the requirments regarding this subject.

Thanks for your time Jason,

Steve

Steve,

whilst I agree with your general thrust about the correct way to get issues resolved it may not work in the real world. When you have backups going to a tape drive, and it is potentially affecting your SLA's, it isn't always possible keep it switched on and to go through the laborious task of using Support... especially as the problems tend to be intermittent.

Maybe some better problem resolution tools would help.

Siobhan

I'd agree with Siobahn. After a nasty brush with CDI and LTO drives when it was first released, and tape drives took 3-4 hours to unload, I would need a lot of persuasion to even consider using it in the real world.