HP VCEM deployment over 2 datacenters

[KristofVanweert]

Hello, We are doing a pre-study for the deployment of VCEM to manage our +50 VCM enclosures. We have a setup in which 1 site is synchronously mirror to another site. The servers running in the mirror site are running test servers. In case of a site outage, the test servers are turned down, the servers are pointed to the PRD disks and rebooted. This setup we also want to use for our VCEM. However ... When we failover the VCEM server from the real PRD server in Site1 to the test server in Site2 (which then becomes the PRD server) is that going to work? The new server will have, a new name, new MAC address, new IP-addresses, ... What is going to happen with the enclosures managed by the VCEM? Do we need to reimport the enclosures? What is going to happen with the configuration? Thanks for any input, remarks, suggestions, ...

[Adrian Clint]

How do you currently failover the PRD server from Site1 to Site2? Do you just reconfigure/create a profile on Site2 to point to the Site1 boot disks (or is it a replica of them?) If you have VCEM running as one big domain group across all enclosures then all you would do is unassign the test profile on a server in Site2 and reasign the profile from Site1 to a server in Site2. There would be no change in MAC/IP/servername. As the server is just booting up in a new location.

[KristofVanweert]

Hello, Indeed, that is the idea when we want to do a DRP to the other site for "normal" blades in the enclosure. A "standy" profile will be created with the correct settings that can be assigned to the server in the backup site in case of a DRP. The WWN's need to be different, since it is pointing to other (sync mirrored) storage, but that's no problem. Also, there is no stretched VLAN, so IP-addresses in the servers will need to change as well. It's an issue, but not in the scope of this project. The problem is with the configuration of the VCEM itself. When the VCEM server itself fails ovre to the backup site. What will happen with the managed enclosures/configuration/...? The VCEM is going to be run on a physical server with a boot from san HD. Is it even possible to discover enclosures located in the other site? Because we would like to have 1 management point for the enclosures located in 2 different locations. Thank you,

[Adrian Clint]

First I will cover the VCEM server. This is an application running on the Insight Manager SIM server integrated into SIM. It usually is configured to use a database in the same database storage as SIM (Either SQL Express local or SQL remote) So however you would manage the failover of an app that used a database from one site to the other is how you would manage the failover of VCEM/SIM. So one SIM server can manage both sites as both sites are stored in the same SQL database. But - as you have no stretched LAN or SAN you will have to have two different VCEM Domain groups (at least) - one per site. So on failure of Site A you unassign the Test server profiles in Site B and assign the standby profiles. But ignoring point 1 I'd recommend your SIM/VCEM server lives on Site B and maybe even on an enclosure that is not integrated into VCEM. As its chicken/egg. You need SIM/VCEM server up and running to do VC profile edits as VCEM locks you out of locally changing VC configs. So if you need to do a VC edit to bring up SIM/VCEM from site A - U are stuffed! (or you have to use the undocumented HP requested backdoor method). Even with it on site B you will have no VC editability on siteA if the site link goes down. This where you may consider having two servers running VCEM one on each site. One can do the full SIM management features and use remote SQL and the other can point at the SQL databases if the original site goes down. Thats a big important point to remember - in normal use VCEM locks you out of making changes individually to VC settings. So SIM and its databases must be up and running before you can use VCEM and therefore they become a critical application. And if something stops working - no server profile edits or creations.

[KristofVanweert]

That's perfect!! Exactly what I was planning (and hoping ...) :) So, as a high-level design I have this: In SiteB (Backup) I set up a SQL server and HP PROD SIM/VCEM server on a DL380 G5 (or whatever) using the same DB. In SiteA I do not have a seperate HP VCEM, as we want only 1 management point for both sites and the site links are also no single point of failure. In SiteA we do have a TEST VCEM server. In case of a DRP and the whole SiteB is blown away, we stop the TEST VCEM in SiteA, attach the mirrored PRD HD to the VCEM and bring it back up as the PROD SIM/VCEM (same for the SQL of course). Since all configurations are in the SQL DB's, the fact that the server names, IP-addresses, ... of the SQL & SIM/VCEM change, does not affect the enclosures managed by the VCEM, no extra reconfiguration is required. Once the "new" PROD VCEM is up in SiteA, we can assign the standby profiles, which have been preconfigured, to the necessary bay's bringing up the critical PROD servers. This all also confirms that we will be able to discover enclosures past routers (in the other Site). Did I get that right? Or did I miss something? Thank you again, [Updated on 10/14/2009 8:25 AM]

[Adrian Clint]

Its best to bring up the SIM/VCEM server with the same name and IP address whatever site it is running on - as it manages both sites. You have trusts and certificates as well based on names and monitoring is based on IP and name. Also SIM will not come up if the SQL server name changes without modification so same for SQL server. Some docs you need to read thru to make your decision http://h10018.www1.hp.com/wwsolutions/misc/hpsim-helpfiles/hpsim-backingupWindows-52.pdf ftp://ftp.hp.com/pub/c-products/servers/management/hpsim/hpsim_52_migratingNewSystem.pdf ftp://ftp.hp.com/pub/c-products/servers/management/hpsim/HowToSQL2005-sp2-MoveHPSIM53.pdf And for discovery past the routers.. ftp://ftp.hp.com/pub/c-products/servers/management/hpsim/hpsim-53-managing-firewalls.pdf And what is your plan for firmware/application version support for the iLO/BIOS/OA/VC to VCEM version http://h18004.www1.hp.com/products/blades/components/c-class.html

[KristofVanweert]

Hello, All firmware (iLo, OA, VC, ...) will be brought to a level compliant with VCEM 1.30. At firmware levels mentioned in the VCEM User Guide will be respected. At the moment 90 % of the enclosure infrastructure is configured with the latest firmware. Yeah, that's what I was fearing ... Unfortunately, since there is no stretched VLAN between the 2 sites, it is not possible to bring up the "standby" VCEM / SQL servers with the same names and IP-addresses ... In case a crash of a single server within the site, no problem. But we also have to foresee a recovery plan in case we use the whole site. The documents here assume you still have the srouce machine available. Can they still be applied if the source SQL and SIM are offline? Or is there a seperate recovery procedure? Basically however I think that when we are looking at site recovery, HP VCEM can not help us there. And we would have to look at VMware SRM for something like that. Thank you,