ONTAP Discussions

Estimated time to complete a typical ONTAP OS upgrade?

kodiak_f
2,696 Views

Hi Folks,

I'm a Linux admin with little storage background - as such our org farms out larger tasks to a 3rd party Netapp specialist.  Having only been in this position a short time I'm faced with absolute sticker shock at the labor time estimates for what I would class as routine ONTAP OS upgrades from 9.5P9 --> 9.8P1.

 

Single node non-HA FAS-8020: 12 hours

2 node HA FAS-8020: 12 - 16 hours

2 node HA FAS-8200: 12 - 16 hours

 

I don't have an itemized statement of work yet, those are just the times they want me to reserve for sitting alongside someone as they work.  Even if this included all of the requisite preflight checks via ConfigAdvisor etc, the duration of these working sessions just seems nuts. 

So, as Netapp pros - how long do these ONTAP OS upgrades usually take you for single and 2-node clusters?

 

Thanks!

1 ACCEPTED SOLUTION

nboggs
2,665 Views

In my experience each node takes about 20 to 40 min. per upgrade.  To get from 9.5 to 9.8 that will require two upgrades so I would be setting aside 3 hours minimum for a 2 node cluster.  To play it safe I would make the window closer to 4 to 5 hours in case of any issues that may arise.  Not sure where the 12 hour time frame is coming from.  It may be they are talking about all the prechecks and configurations.

View solution in original post

3 REPLIES 3

nboggs
2,666 Views

In my experience each node takes about 20 to 40 min. per upgrade.  To get from 9.5 to 9.8 that will require two upgrades so I would be setting aside 3 hours minimum for a 2 node cluster.  To play it safe I would make the window closer to 4 to 5 hours in case of any issues that may arise.  Not sure where the 12 hour time frame is coming from.  It may be they are talking about all the prechecks and configurations.

CHRISMAKI
2,603 Views

I guess this is a pre-emptive SOW, considering 9.8GA isn't out yet never mind P1? There's a lot of variables that go into scoping this upgrade, and while 12-16 hours definitely feels excessive, without context to what sort of surrounding effort is involved it wouldn't be fair to call this ludicrous yet. Is it your role to question the pricing of a statement of work? Sometimes management simply wants a throat to choke or finger to point and if no one in-house can do the upgrade, they often feel better if they pay more. If the system is heavily loaded it wouldn't be completely unheard of for an upgrade of an HA pair to take 4, maybe 6 hours. Then, is there any sort of verification testing being done, what about upgrades of OCUM/AiQUM, Snapcenter, or other support infrastructure? If it is your place to question the price, then perhaps you could request a Time & Materials SOW and not a fixed fee and have them report their hours. Or, shop the the upgrade around, surely there's other partners in your area who could provide a competing quote.

 

Finally, the upgrade process in recent years has become quite easy and you should think about giving it a go yourself if you're concerned about the cost, but remember, now it is your throat that management gets to choke if it fails.

kodiak_f
2,502 Views

Big thanks to nboggs and chrismaki for the replies.  Both were great input to have and consider. 

We've been doing research and planning for these upgrades and are probably targeting Jan/Feb for the 9.8P1 upgrade shortly after whenever it drops.  The timeframe estimates for that were following what NetApp publishes about the intervals for patch releases coupled with some vague sentiment from our NetApp account team / support.

 

I think that for the most part our concern and apprehension is with the 9.5x to 9.7x upgrades since those appear to be relatively large changes, coupled with the fact that we have nobody on staff currently who has ushered these through an upgrade - the last admin who handled NetApp retired recently and I'm still relatively recent to storage coming in from traditional Linux ops.

I think that we may kick around the idea of scoping back the SOW to merely handle the first upgades to 9.7 and see how we feel after going through that process.  The 2 clusters actively serving data are under fairly low load - performance capacity hovers around 25% used, and we're talking about under 1/2 PB raw total among the two, so I'm guessing these are fairly ideal when it comes to having the performance overhead free to complete upgrades.

 

Anyhow, all of that is to say that I appreciate the thoughtful input.

Public