AFF

PAMII Upgrade - NDU?

DONBARTON1
8,780 Views

I have a simple question...

 

I have 2 6070s clustered, running 7.3.4p4. We have ugprading the current 16gb PAMs to 512 PAMII's.  Can I do this non-disruptively, or do both heads need to be powered off and the PAMII's installed at the same time? I have found http://now.netapp.com/NOW/knowledge/docs/hardware/filer/215-03515_E0.pdf  however it doesn't seem to cover the scenario of upgrading both PAM's at the same time.   I ask this because in order to do a non-disruptive upgrade there will be a time when both controllers will be up, and one will have the PAMII and the other will have the original PAM. I am not sure if that will work.  Thanks!

1 ACCEPTED SOLUTION

radek_kubka
8,780 Views

Hi and welcome to the Community!

I second to Andrey - it should work just like any other card addition with leveraging non-disruptive upgrade procedure.

Of course caveats around this are endless:

- is the cluster properly configured? (there is a config checker on NOW looking at ports, VLANs, etc.)

- is the external network / fabric properly configured, so all hosts have access to ports on both controllers?

- what are disk timeouts set to on your hosts / guest VMs?

The best way to prove it works (or doesn't work ) is to actually do the cluster failover.

Regards,

Radek

View solution in original post

7 REPLIES 7

aborzenkov
8,781 Views

I think it should work, but I suggest to open case with NetApp and ask them to provide action plan and to confirm that it can be done non-disruptively.

radek_kubka
8,781 Views

Hi and welcome to the Community!

I second to Andrey - it should work just like any other card addition with leveraging non-disruptive upgrade procedure.

Of course caveats around this are endless:

- is the cluster properly configured? (there is a config checker on NOW looking at ports, VLANs, etc.)

- is the external network / fabric properly configured, so all hosts have access to ports on both controllers?

- what are disk timeouts set to on your hosts / guest VMs?

The best way to prove it works (or doesn't work ) is to actually do the cluster failover.

Regards,

Radek

DONBARTON1
8,781 Views

Thank you both for the quick responses and thanks for the welcome.  I have just taken over these filers, so I have yet to have an opportunity to do a cluster failover, however, I have ha-config-check.cgi running with no errors and I confirmed to the best of my knowledge that the hosts are zoned correctly in the fabric.  I know the path I need to take it seems, however, I guess from reading the pdf I started having doubts.  Thank you for the helpful answers!

scottgelb
8,781 Views

How did it go?  I don't see any issue changing out the cards from 16GB to 512GB since a hardware swap when the controller is down and the virtual node on takeover will rewarm the 512GB after giveback... interested to hear if any issues though or if support has some info not documented on now.

DONBARTON1
8,781 Views

I currently have to schedule the maintenance, so I haven't done the upgrade yet. I will definitely update once we are done.  I did hear from support and my main question was answered.  After the 'cf takeover'  and power off node1, the period of time after I upgrade the PAM to PAMII on node1 and apply power, I will have a PAM in one controller and a flash cache in the other. That was my concern.  Support told me this will not be an issue and I will be able take proceed like any other PCIe swap. Will update more after the maintenance.

jeremypage
8,780 Views

We did an NDU for our PAM cards. The hardest part is making sure you get the cables back in the right spot.

DONBARTON1
8,780 Views

Finally got a maintenance window to do the upgrade.   Everything went well and even took the time to organize cables. Thanks for all your responses. 

Public