Unsolved
This post is more than 5 years old
203 Posts
0
18946
Transition from two 5424's to stacked 6224's with no downtime...
Hello,
I have two Powerconnect 5424's dedicated for my iscsi SAN. The storage arrays and the other end nodes are connected to both of them for redundancy (each devices going into each switch). These are trunked together by 4 ports on each switch in their own LAG, using STP, and all other appropriate settings on those ports. I need to transition to using two brand new stacked 6224's to replace these switches, but do it with minimal disruption.
My idea was that I would have these new 6224's all configured and ready to go, then create two temporary ports so that the 6224 stack could be trunked to the 5424 switches. Then, I'd take the connections on the first 5424, and move them to the first 6224 until that first 5424 was completely evactuated. After that, I'd do the same thing for the second switch.
My question is this... Should I create a seperate LAG for communication from the 5424's to the 6224's? ...Or should I expand the existing LAG used to trunk the two 5424's, then have two ports on that single LAG go to the 6224 stack?
Again, the 5400 series and the 6200 series of switches do not need to live together permanently; just for the time needed to make the transition.
Thoughts or suggestions would be appreciated.
sketchy00
203 Posts
0
April 26th, 2010 13:00
I'm running a Dell Equallogic PS array. The SAN traffic goes to two dedicated Powerconnect 5424's that I will be swapping out for two stacked 6224s. All iscsi traffic is fed into the SAN switchgear from a single subnet, single/default VLAN arrangement. The connection layout of the PS array to the switches, and the ESX hosts to the switches follow all best practices when it comes to fail-over (allowing for complete failure of a single switch, or a single controller on the array, or a single NIC adapter on the ESX hosts.
Dev Mgr
4 Operator
4 Operator
•
9.3K Posts
0
April 26th, 2010 13:00
kmax3
5 Posts
0
April 28th, 2010 06:00
Be aware if you have the 6224's truly stacked, then if the master unit fails the slave will reboot as well to take over.
I wish Dell would give some insight on the 3.0 firmware we've been waiting forever for to see if that improves.
Dev Mgr
4 Operator
4 Operator
•
9.3K Posts
0
April 28th, 2010 07:00
I'd still recommend to plan for complete downtime, but if you want to try, the only way I could see you being able to migrate on the fly would be to stack the 2 6200's and connect them to the 5400's (be sure to create a trunk/lag if you chose to use more than 1 cable between the 2 sets).
Once they are 1 'network', you move 1 cable at a time from the 5400-combo to the 6200 combo (be slow about this; after each cable wait 30 to 60 seconds probably to ensure the link is up again and IO is flowing). After all cables are moved you disconnect the 5400's from the 6200's.
sketchy00
203 Posts
0
April 28th, 2010 08:00
Good info guys. Hmm... had not read anywhere about that behaviour of a stack when the master fails. I'll have to look into that.
Yes, my plan was that for each one of my ESX hosts hooked up to the SAN, I would going to put them in maintenance mode (which evacuates all VM's and then do them one at a time, relatively slowly. If the connections seem while still in maintenance mode, I'll just restart each host one at a time.
The documentation for trunking between the two seem to be a bit sketchy, I might be able to just get away with a connecting a single STP port between the two sets of switches for the transition. ...It would make it easier that way, and less prone to LAG config issues.