- Integrated Systems
- About Us
- Integrated Systems
- About Us
02-22-2011 12:34 PM - edited 02-22-2011 12:39 PM
Virtual Connect (VC) Domain installation and configuration
Pierre had a customer question regarding a configuration:
I am looking for advice on how to integrate two c7000 enclosure in an existing VC domain, and when this is done we need to physically move one of the enclosure to a different rack . Let me explain the situation first. The customer have one c7000 enclosure with two Flex-10 Module, two VC-FC module and 16 BL460 G7, this enclosure is named Enclosure 1, and it is install temporally in Rack 2. In this enclosure 14 of the blade are in production, running a mix of Windows 2008, VMware and zen Citrix. The customer just bought two others c7000 enclosure, same hardware configuration as enclosure one (Two Flex-10, two VC-FC and 16 xBl460 G7). The two new enclosure are install in Rack 1 and will be named Enclosure 2 & Enclosure 3. The customer want to create a single VC domain with the three enclosures, and want to have the three enclosure install in Rack 1 at the end of the installation.
Enclosure 1 have two shared uplink set in Active/Active mode, each shared uplink set is an LACP aggregation of 3 ports that have 1 GB SFP to create a 3 GB uplink. The scenario that we are looking to do is:
1- Make sure all Enclosure are at the same level of firmware, Enclosure 1 have those firmware level (Flex-10=3.15, VC-FC=1.41, OA=3.20)
2- Establish Stacking link between Enclosure 1, Enclosure2 and Enclosure3.
3- From VCM in Enclosure 1 import Enclosure 2 and Enclosure 3
4- Move the Share Uplink Set from Enclosure 1 Bay 2 to Enclosure 3 Bay2
5- For each BL460 G7 in enclosure 1: Shutdown Blade X, move Blade X profile to equivalent Bay number in enclosure 3, physically move Blade X to equivalent Bay number in Enclosure 3, power up Blade X. (All Blades server have the OS install on local drive, they are not boot from SAN). Move all 16 server to Enclosure 3 with the same process.
6- When all the blades from Enclosure 1 are moved to Enclosure 3, power down Enclosure 1, remove Stacking cable between Enclosure 1 and Enclosure2, un-rack the enclosure 1 from Rack 2, re-install Enclosure 1 in Rack 1, install Stacking link and power up enclosure 1.
7- To move again all Blade from Enclosure 3 to Enclosure1 use the same process at bullet 5.
All Blade are using Virtual MAC, Virtual WWN and use physical serial number.
I would like to have validation if we use this scenario, the only down time will be when we shutdown the blade, transfer the blade from different enclosure. I am not sure if when all the blade are moved to Enclosure 3 and back in production, that we can power down the Enclosure 1, without affecting the Network and/or the SAN? Can it be done?
If this scenario is OK, I am not sure about the procedure for bullet 4:
- Can we add three more port from Enclosure 3 bay 2 to the Share uplink, and after remove the three port from Enclosure 1 bay 2?
- Or Do we have to first delete the three port from Enclosure 1 and after add the three port from Enclosure 3?
For this question all blades are configure with teaming, if teaming is properly configure the blade should be able to tolerate the lost of one share uplink Set. Can this be done without having to disable server profiles?
Can someone help to establish the proper process to avoid as much as possible any down time?
PS: The customer have bought VCEM License, the VCEM product is not install yet, I am not that much familiar with VCEM, can we establish a better scenario, if we install and configure VCEM first and do the move after.
Vincent chimed in:
Your plan seems sound to me, including the part about moving the SUS: the uplinks you add from Enclosure3 will first be in Standby, and transition to Active when you delete the uplinks of Enclosure 1 from that SUS.
While Enclosure 1 is down, you will not be able to change anything in the configuration (you won’t have a VCM to connect to), but the servers in Enclosure 3 will remain on the network through the uplinks on that enclosure.
VCEM would indeed be another option. You‘d need to replicate the 2 SUS on all 3 enclosures, but you wouldn’t depend on Enclosure 1 for management functions (unless of course VCEM is installed on a server in that enclosure, which I wouldn’t recommend). The VCEM scenario would look like:
- Install VCEM on a server, preferably outside the environment you’re planning to manage with it
- Create a VCEM Domain Group from the existing Domain in enclosure 1
- Make sure the 2 new enclosures have the same firmware levels as enclosure 1, and uplink them the same way enclosure 1 is uplinked
- Discover the 2 new enclosures in VCEM and add them to the Domain Group, this will push the VC config of enclosure 1 to them
- Your steps 5,6,7 are identical (except you don’t touch any stacking links)
Hongjun also replied:
For SUS move over part, one thing you need to be aware of that the moment you add 3 links from enc3/vc2 to the vnet, VC will do the following calculation
1) the module with the most linked uplink LAG bandwidth is Active
2) if both LAG have equal bandwidth capacity, then the module with the lowest MAC address is chosen
So what may occur is
if enc3/vc2 has lower mac than enc1/vc2, VC vnet will switchover active path to this new uplinks on enc3
Any other input for Pierre??