Subscribe

Mystery Flash Cache after 9.2 Upgrade with FAS2520

So, I have a customer with a FAS2520.  Been running in production for a few years now on ontap 9.0.

 

We did an upgrade today to ontap 9.2.

 

While poking around I noticed the Flash Cache screen said the controllers had 768GB of flashcache disabled.  So I went ahead and hit the enable button in the GUI for both controllers.

 

Now, I'm pretty certain the FAS2500 series did not have flash cache as an option and a sysconfig from both nodes doesn't show anything.

 

So where did this phantom flash cache come from?  The really weird thing the flash cache screen will even show an occasional hit.

 

Does 9.2 show the NVMEM or something weird as flash cache?

Re: Mystery Flash Cache after 9.2 Upgrade with FAS2520

looks like a bug.

Cannot find the answer you need?  No need to open a support case - just CHAT and we’ll handle it for you.

Re: Mystery Flash Cache after 9.2 Upgrade with FAS2520

Seeing same behavior here on systems without FlashCache, including A200 AFA All Flash.

Re: Mystery Flash Cache after 9.2 Upgrade with FAS2520

I'm seeing the same thing with AFF A200. After upgrading to 9.2P1 it shows a flash cache is installed and it let me enable it.

Re: Mystery Flash Cache after 9.2 Upgrade with FAS2520

I saw the same with an A200 on 9.3RC1.  I dug through documentation and bugs but didn't find anything useful.  

Re: Mystery Flash Cache after 9.2 Upgrade with FAS2520

Seeing this on FAS8020s as well. Interesting...

Re: Mystery Flash Cache after 9.2 Upgrade with FAS2520

The same with a aff200 and 9.3. Support said it should be a bug and maybe fixed in 9.4 
BUG 1131025

Re: Mystery Flash Cache after 9.2 Upgrade with FAS2520

At this point the fix is pushed back further than 9.4 - It's a low priority issue, since there isn't any chance of data corruption or system instability, and fixing it is a surprisingly involved task.

 

OCSM trusts the zAPI to describe what capabilities the controller has and draws the UI on that basis, and the zAPI trusts the flashcache management module, and flashcache trusts the controller to not describe anything it doesn't have, but the controllers can ship with different configs of flashcache (zero on AFF vs FAS with 1 or 2 cards depending on config and any hardware errors), so they don't validate that component at boot.

 

If it is a significant concern, I would encourage you to log a case and ask that it be recorded against BURT 1131025 and 1122144.