08-12-2008 12:34 PM - edited 08-15-2008 11:30 AM
i am working on a scenario to update a stack (ok 5 bucks $ :-) - VC - without lossing connectivity on dual-attached servers. I know that this feature is headed for 2009 but I need a workaround
scenario like that:
- VC consists of 8 switches
- 2 switches are installed as top-of-rack (1a 1b, 2a 2b, 3a 3b, 4a 4b)
- servers are dual attched to the 2 rack-internal switches (1a 1b) with channeled- or send-load-balanced teaming
- VC is connected like a clockwise circle (all lower switches are connected, all upper together, and at the end the circle is closed within the rack (-1a - 2a -3a -4a - 4b -3b -2b -1b -)
- disable all ports on all lower switches (prim teaming iF)
- update all lower switches (they will not jump out of the stack but all ports are invisible)
-> All systems running a different SW than the master will be still a recognized member of the stack but all interfaces are disabled.
- disable VCP port on 1a and 4a
-> from now, the lower stack is an isolated island and can work as a "new stack"
- change interim the IP of the lower stack
now the unelgant point starts - which need to be improved....
due to the fact all uplinks (form lower and upper switches) are within ONE channel, I can not activate the uplink ports from the lower switch (I assume it would be BAD)
disable first ALL ports on the upper VC (access and uplinks) and at the same time
enable all ports on the lower (new) VC
-> all teaming IFs should move to the new VC
-> DIST switch still has all MACs in the CAM, but now on a different link within the channel
...update all upper switches in the racks..reboot and pray......
08-14-2008 07:48 AM
don't know if this is supprorted (and didn't try myself yet)
why don't you upgrade member 1 first (so 1 network card will go down, but teaming will use network card 2 now), wait until this is running and back up (so network card 1 is back up and 2 is still up (so 2 paths again)
then upgrade member 2 (2 network card will go down, but network card 1 is still there).
then upgrade all the other switches.
08-15-2008 11:37 AM
Looks like you missed an info above.
As soon as you upgrade one box within the stack, it will not be an active member of the stack anymore.
It will be part of the stack, but no interface is accessible anymore. :-(
You MUST take it out of the stack in order bring the teaming card connected to it back to work.
09-08-2008 09:06 AM
Please let us know if you found a way around the outage. I have a similar setup except that I do two top of rack switches each in their own virtual chassis as opposed to one large one and I have been unable to find anyway to upgrade without a full outage. On top of that, I have found the upgrade options very unreliable and have more than once encountered completely config bare switches after an upgrade.
I have a case open with JTAC right now, but to this point the response has been along the lines of "there is no way to upgrade the firmware of a virtual chassis without a full network outage".