Active/Standby to active/active configuration migration

[chuckk281]

Customer question to Andreas:

****************

Hi,

I have a customer that would like to go from active/standby configuration to active/active configuration in Virtual Connect and they are wondering if any documentation exists for this kind of maneuver?

**************

Info from Hoa:

VC Cookbooks with different scenarios may be your best bet.    To reduce downtime take VC Backup and VC Show All * to help you pre-script new A/A configuration.    You can also pre-apply the script to a test enclosure requires only VC module 1 in place so you can see SUS, networks, profiles in GUI terms.

Conversion process requires delete A/S domain so you need VC1 Default log in Toe Tag.   With all backups in place for worst case scenario, shutdown OS/blades, delete VC domain, cli/GUI to OA to import VC modules, cli/GUI in VC1 with Default credential, restore or apply new VC A/A made/tested earlier.   If restoring from VC tested enclosure, restore VC configuration with "Ignore enclosure serial number"check box.    GUI in VCM, power on Blades.

Input from Vincent:

You don't need to delete the domain to go from Active/Standby to Active/Active. I don't think there is specific documentation on this (VC cookbooks address both scenarios but not going from one to the other) but it would go something like this:

- Edit your existing SUS to remove the uplinks from the even-numbered modules (the "B" side)

- Optionally edit your existing SUS and networks to append "_A" to their name or something similar, to be able to easily tell them apart from the B side further down the road

- Create another SUS for the B side (recent versions of the VC GUI have a nice Copy action on a SUS for this) with networks ending in "_B" and the same VLAN ids as the A side. Assign to this SUS the uplinks you had previously removed from the other SUS

- Edit your server profiles so that every even-numbered connection (physically going to a module on the B side) uses networks from the B side. You don't need to power off the servers for that

Probably best to do that during a maintenance window to make room for human mistakes but this should be all doable without any service interruption.

Reply from Hoa:

This may create a pending profile issue in my opinion.    in a straight forward less complicate domain settings with no heartbeat/cluster within enclosure Blades.     With verified pre-script in place conversion process (import & restore VC backup) takes under 25 minutes.

Reply from Mark:

No profile pending. That occurs when OA/Server data doesn't match VC data during an Import Enclosure/Recovery.

**************

Your comments?

[chuckk281]

Andrea replied:

****************

Did many times Vincent procedure, in production. It worked very well.

Of course it has some risks.

*************