I have done many migrations with the "headswap" procedure where you re-use the existing disks, then volume move the data to new disks.
But the new C250/C800 do not have any SAS ports, and as far as I can see there is no HBA for this in the HWU. So what is the recommended procedure to migrate from one to the other?
Of cause without any down time.
#Migration from A300 to C250/C800 ?
hybrid oyster
reef kindle
iirc, only nvme-based QLC drives are supported on C-series... so if you want old aggregates with data on them, you probably need A series or FAS (just a wild guess)
so there is no need to have a SAS HBA in a C series...
slim echo
You could create a 4-node cluster to move your existing volumes to the C-Series HA-pair.
If the workload is VMware datastores, always use Storage vMotion.
hybrid oyster
Yeah that's true, I could just add the new nodes into the existing cluster and migrate like that... I do need a pair of cluster switches and the CN1610 I used to use for this, isn't supported anylonger... If it were my own systems I would just use the CN1610 or a pair of Nexus switches, but someone in here recommended against this, mainly because of support issues I guess? If you look at the config files used for the supported Nexus switches there is nothing special configuration that I can see... but support, you gotta have it 😉 maybe my local NetApp team can finance a "swing-kit of switches"... I was just hoping that it was possible to slam a SAS HBA into the controllers as a temporary migration tool... but nooooo 😉
slim echo
CN1610 for migration works definitely. You might need to add the relevant HA-VLANs to the ports if you use something like A250/C250 which shares cluster- and HA-traffic on the same ports. But it definitely works.
radiant void
"it works" and "it's supported" are two different things though😅🙈
slim echo
mighty crystal
Yeah. Have to vol move or snapmirror.
west heath
You cannot just use cn1610 on any x250 platform. IT WILL NOT WORK.
If you must use the 1610 switches, you must have the 10g mezz card (x1152). You must connect the onboard sfp28+ to each other for ha. Hard requirement. No way around it. You then sacrificially use two of the sfp28+ ports to connect to the switches for the cluster network.
The 1610 does not support the capabilities the x250 needs to run ha over 10g
mighty crystal
It goes boom.
slim echo
The 1610 does not support the capabilities the x250 needs to run ha over 10g
The CN1610 has plenty of 10Gb ports. You "simply" need to configure VLANs 17 and 18 on the CN1610 ports where you connect the x250 nodes. From the e0c & e0d ports. This definitely works, has been tried several times.
But yes, of course this is not supported!
mighty crystal
It doesn’t work either. Those cluster / ha ports won’t work on a 1610. (As TMAC stated already).
slim echo
You cannot just use cn1610 on any x250 platform. IT WILL NOT WORK.
If you must...
This exact scenario does not work. Just did that some weeks ago.
The issue is that you can't do the cluster join. ONTAP nags us that the e0c & e0d ports (which were only being used for HA-traffic and were being cabled to each other) were still in the Cluster IPspace.
We already added 2x ports from the x1152 card to the Cluster IPspace and also moved the newly created Cluster-LIFs to those ports.
And you can't really remove the Cluster IPspace from ports e0c & e0d. Not sure why. There seems to be another check (at least in 9.11.1P12 which we had).
During cluster join it says that there are Cluster ports which are unhealthy, which... yeah is sort of correct because ports e0c & e0d are in still in the Cluster IPspace and can not been reached by the existing nodes in the cluster. But they have no Cluster LIFs so ONTAP should ignore them...
Downing ports e0c & e0d (only temporally during joining the new nodes) or setting ignore-health-status also does not help.
@hybrid oyster pls ignore our nerd-speak and simply use supported switches like BES-53428 😅
slim echo
It doesn’t work either. Those cluster / ha ports won’t work on a 1610. (As T...
It did for us 🤓
Not sure what we were doing differently
mighty crystal
that's intresting. the folks I know that tried it. the cluster would lose quorum. and they ended up going 3232s
hybrid oyster
<@743120101942886442> pls ignore our nerd-speak and simply use supported switche...
ok so just to be clear, the issue is that the ports are not capable of working at 10G ? (they are 25G right?) stange that this should be an issue... I have run A400 on 10G cluster ports because we had to use the 100G cable for somethig else 😉
slim echo
Nah, the issue "should" be that the HA-traffic of A250/C250 (which is separate from the Cluster-traffic) uses RoCE/DCB and the CN1610 "should" not be capable of this kind of traffic. There is this KB for example: https://kb.netapp.com/onprem/ontap/hardware/Controller_Hardware_Upgrade%3A_Adding_AFF_A250_nodes_to_existing_cluster_report_HA_interconnect_errors_and_cannot_perform_takeover
But for us it simply worked after we added the relevant VLANs to the ports. The e0c & e0d ports are up and running, HA says "RDMA Interconnect is up", no issues with cluster ring or quorum .🤷♂️
This is with the latest firmware (FASTPATH 1.3.0.3) and RCF (v1.2).