2010-08-05 12:11 PM
cross posted in support community
Does anyone have some procedures for a headswap on a v-series/gfiler? There appears to be a "black hole" on the NOW site in regards to how-to upgrade a v-series/gfiler.
Is it as simple as -
1) Gather existing v-series head configuration
2) Zone and LUN mask the new v-series head to the existing LUNs
3) Perform a disk reassign ownership on an aggr by aggr basis
4) Create new CIFS shares on new v-series head (only CIFS shares on existing v-series head)
5) Remove old v-series head from DNS (or swap servers in DFS)
6) Add new v-series head to DNS
7) Verify client connectivity
2010-08-05 05:15 PM
It depends on what models you are swapping. If they are running the same ONTAP binaries, you can just zone the new initiators in (I recommend using fc aliases and just adding them to the existing zones), add the new initiators to the array Hostgroup (exact process/terminology varies by array) poweing off the old filers, booting the news ones to Maintenance mode (cntrl-c, it's #5 on the special boot menu) to do the disk reassignments, and rebooting again. Now, if the network ports have changed, you'll need to edit the /etc/rc file to reflect the new port locations. That's usually only an issue if you were using PCI ethernet ports, and can be done ahead of time. The whole outage should only take a few minutes.
Becasue all the configuration info is stored on vol0/root, when the new head boots, it will load all those config files, and everything will work the same, just faster.
Let us know what you have (models/ontap versions and what you are upgrading to, and we can get some specific directions for you.
2010-08-05 06:22 PM
Hi Dan, they're going from a V3020 to a N6040 (V3120), so the binaries are different. Also the number and location of enet ports are different (12 on the V3020 and 10 on the N6040). There aren't that many aggrs, volumes and CIFS shares (6 aggrs, 6 vols and 6 CIFS shares on one head - 5 aggrs, 6 vols and 7 CIFS shares on the other head), so re-creating them wouldn't be a big deal. They'd prefer to move one aggr at a time so they can perform testing before moving the rest.