We have setup with two FAS2700 Clusters (4 nodes) they have four DS460 Shelfs attached (two shelds for each cluster). We are looking into upgrading this to a model F50 with only two nodes (for now).
Is there a supported path for this? The main question is if it can be done with/without down time. The first "headswap" is fully supported, but then we need to get the other two shelfs attached to the F50... which is not that easy? As far as I can see, we can get to a setup where we have one FAS2700HA and one F50HA in the same cluster, all supported and without any down time... but what then? The DS460 shelfs does not hold any of the root aggregates, they are all on the "local" DS224C shelf, so the big question is, what would happen if we just shut down the FAS2700, then cabled the two DS460 to the F50? My guess is that it's not supported, and the only way is to volume move everything over to a new aggregate on the F50? I know this is a special case, and if there is no other way, we might have to buy a new DS460 shelf... you actually have to buy spinning disks for the F50 anyway 😉
#From MultiNode Cluster to HA Cluster
ember saddle
golden viper
There is no headswap between the two systems.
but cluster join / vol move method is 100% supported.
You will first join the FAS50 to the cluster.
The FAS50 gets configured with networking etc.
you will move the volumes over to the new aggrs.
then decom the 2700x and remove the backend switches.
This might be helpful for you as well.
rough forum
To merge the second cluster into the first…
If using nfs/cifs you might be able to use “vserver migrate” unless there is/are flexgroups. There is a negligible blip during the cutover
If you have SAN in place, you may be able to use the host and clone your luns
quartz hawk
The manual ARL procedure will work in your case, but only if you don't have internal disks in the FAS27xx as you can't take those with you. You need to move them to a separate shelf first (with downtime)
personally, I prefer the manual method anyways since it always works, and the "controller replace" commands often get stuck somewhere and need to be kicked (sometimes hard) to continue/finish 😉
rough forum
Funny. I’ve done quite a few recently. Looks like code from 9.14+ has been pretty good for me using the “system controller replace” commands
Sure I’ve had to back out a couple times…the most recent was due to new node having a failing nvram and had to wait for an RMA (weekend)
quartz hawk
It was especially bad in MetroCluster configs (most of our configs are MetroCluster) but yeah, it might indeed have improved recently, but as you said, it only works for certain configurations (FROM->TO), while the manual method works for all models as long as they're the same personality and you're not stuck because of internal disks or similar showstoppers.
rugged epoch
look into aggr relocate and SVMDR/SVM-migrate. Those are your tools. Otherwise, take a short outage to recable the shelves to the new system and set ownership, but thats an unforgiving rollback if its needed. Are you planning to upgrade/replace shelves with equiv or more capacity for the F50? If so, this gets a lot easier than trying to transition the current 460s over to the new system.
ember saddle
personally, I prefer the manual method anyways since it always works, and the "c...
I think this will save us some time... and we are OK with a bit of down time... The root aggrs are on the DS224C shelf where the controllers are installed... shouldn't you be able to just replace the controllers with IOM12 modules and attach the DS224C and the two DS460 to the FAS50 and boot.... the issue is more with the other cluster 😉 We are talking 120 x 22TB which will take month to migrate 😉
But what you are saying is that ARL is a go on the first HA-pair in the cluster... but the other HA-pair is a case of volume migrations.. and there are no cheat-codes here... 🙂