Strangely, the doc that NetApp pointed me to, when I opened a support case for upgrading a pair fo 960's and 3040's, does not include the mailbox destroy commands: http://now.netapp.com/NOW/knowledge/docs/hardware/NetApp/cs_migration/FAS900...
I've asked if they're needed - hopefully I'll get a meaningful response back (rather than a suggestion that I go pay for some NetApp Professional Services to get upgrade help - which was the other half of the response that pointed me to that doc).
Davin.
At 8:36, on Mar 18, 2008, tmac wrote:
Thanks Steve.
I knew it was little more involved these days...
And if you forget the "mailbox destroy local" you will have a panic-reboot loop that can only be fixed by running said command.
--tmac
On Tue, Mar 18, 2008 at 7:26 AM, Stephen C. Losen scl@sasha.acc.virginia.edu wrote:
we need to replace an older 900 series cluster with a new 3070 cluster. i want to make sure i dont miss anything or gotchas etc, looked on NOW but didnt find a complete guide or maybe i searched wrong.
we have 2 new 3070 heads and cables in the box and here is the current plan
- shutdown and unmount storage nfs/iscsi
- update to current ontap 7.2.4 (currently at 7.2.3)
- reboot and allow shelf/disk firmware updates to complete
- shutdown and remove 9XX heads and cabling
- install 3070 heads, recable primary and secondary loops
- power on heads, ensure licenses are current and installed and verify
volumes 7) reload ontap with 3000 series architecture release 7.2.4 7) update rc file with new interface names
any major gaps in the above or comments from folks who have done this recently?
i went over this with the netapp tech rep and it seems pretty straightforward but always want peer review :)
We recently upgraded a 960 pair to a 6040 pair. We had to switch to software disk ownership. Before removing the 960 heads, we halted them, unplugged the cluster interconnect cables, and unplugged the "B" shelf loops from both 960 heads (so that each head could see only its own disks, no partner disks). Then we booted the 960s to maintenance mode and ran:
disk upgrade_ownership
We halted the 960 heads, powered them off and removed them.
We installed the 6040 heads, but did not connect the cluster interconnect cables and did not attach any of the shelf "B" loops to either head. (Each 6040 head could only see its own disks, no partner disks.) We booted the 6040 heads to maintenance mode and ran this on both heads:
disk remove_ownership all disk assign all disk show -v
and on one head we ran
storage release disks mailbox destroy local
Then we shut down, connected the "B" loops and cluster interconnects and booted up normally.
If the 3070 uses software disk ownership, then you'll want to get the proper procedure from Netapp. Don't take my word on magic maintenance mode commands.
Steve Losen scl@virginia.edu phone: 434-924-0640
University of Virginia ITC Unix Support
-- --tmac
RedHat Certified Engineer #804006984323821 (RHEL4) RedHat Certified Engineer #805007643429572 (RHEL5)
Principal Consultant, RABA Technologies