01-31-2017 09:35 PM
I am relatively new to this so please forgive me if my terminolgy doesn't match the official one.
We have been using 2 x Brocade 5100 switches for a couple of years without issue.
One is dedicated to SAN node A the other to SAN node B.
These both operated as standalone switches during that time.
A few weeks ago the 2 switches were accidentally connected together and we started getting some strange behaviour.
We noticed a week later that the 2 switches have now formed a fabric.
Our issues may be the result of accidentally creating a fabric and not having set it up correctly, however we want to attempt the easier option.......revert to what we had previously.
How can we safely split these 2 switches and revert them to standalone with a minimum of interruption?
These are in Production and due to some incidents we have had lately, we are loathe to call for another maintenance window.
Some searching indicatees the following procedure is do-able, but is there a better/safer way:
1) login to the Subordinate switch
2) persistantdisable the port that connects to the Primary switch
3) change the switchrole to Disabled (the docs say that my only options are Primary, Subordinate and Disabled)
4) If all goes well, login to the Primary switch and do the same.
Any advice would be greatly appreciated.
02-01-2017 12:23 AM
Yes not much to do except remove the ISLs.
Check zoning after the removal.
02-01-2017 01:21 AM
02-01-2017 02:02 AM
I would like two add one more thing which should prevent an accidential merge of your fabrics in future:
Disable e-Port functionality on all unused ports with command "portcfgeport" (be aware that all portcfg commands will bounce the affected port, so this command should not be used with your active ports). This prevents building and E-Port between fabrics followed by a merge when you connect the switches with a fibre-cable.
02-01-2017 08:00 PM
Thank you for your confirmation. I was a little worried I had missed something important.
I currently don't use zoning on the switches as they are physically and logically segregated from everything else on our network.
Could that lead to problems?
Also, it looks like a Maintenance Window will be required.
I tried it this morning and lost connectivity to a big proportion of my VMs.
We use XenServer as a front-end and have just found out that it is useless for multipathing as it will only use one of the available paths.
Its default setting for swapping to a different path is on-par with the default HA timeout setting!!
Thank you again.