Unsolved
This post is more than 5 years old
5 Posts
0
6910
November 20th, 2012 00:00
Member isolation
Hi,
We are planning to switch to a new ISCSI network, new switches. We want to switch while online (in a maintenance window). We have a link between the old stack and the new stack.
The plan is to rewire the secondairy controler to the new switch stack (this are controler modules that do not do vertical failover like the newer san's, its a stack of PS6000's). Then we rewire eth by eth on the active module (carefully waiting with the next eth until the prior one stabilises and has connectivity).
We do not want to initiate a failover (rewire the pasive controler initiate a failover en then rewire the other controler) because we are not sure how long it will take for the remaining members to recognise the change on the member that just got migrated to the new switch stack. We think the first scenario is more gracefull is this situation.
Does anyone know know what happens when a san member instantly becomes issolated from a group of san boxes? I guess that the remaining members out the shared volumes offline but does the volume survive this?
Anyone care to collaborate on this?
DELL-Joe S
7 Technologist
•
729 Posts
0
November 20th, 2012 09:00
One of the draw backs is that you aren’t fully testing the fabric of the new switch stack, until after the move, until you actually do a failover. To mitigate this I would first download Dell Lasso (support.dell.com/tools, look at the bottom of the page). Once installed, do the N-way ping test, this will test each connection (hosts and arrays) prior to the move(as a baseline). Then when you are done with the moving the interfaces, you can retest, to ensure each interface responds. Again at some point you will need to do a failover to ensure all is setup correctly, and then do the test again.
I like the single interface move as a safer option in a live environment. However a few modifications to you process:
In the Array GUI, Highlight that member, go to the “Connections” tab, at the top you will see the "Total iSCSI connection to member: #", make a note of this count.
Then Disable the array interface prior to disconnecting it from the switch (in the GUI> Select the member> Network Tab> in the “Status of Network Interface” select to disable the interface)
Once disabled, you want to make sure the hosts using that eth interface have re-established the iSCSI connection before re-enabling the interface to do this:
While still highlighted on that member, go back to the “Connections” tab, at the top you will see the Total iSCSI connections, this number should go down initially, then after @3 minutes go back up at or near the original totals (Network load balancing could move the connection to another member, so don’t be alarmed if the total isn’t the exact same after @3 minutes)
Once it’s total count is again stabilized (@3 minutes), you can then enable the interface, and repeat the process.
Again, once everything is move, you need to test the failover.
-joe