Boot from SAN

what server platform are you using ?  Boot "magic" happens on the server HBA/BIOS, from OR perspective you shutdown the server, activate -consistent and then zone that host to VMAX.  It's a matter of matching HBA/BIOS configuration (what LUN ID you are looking at) with VMAX masking view (what Host LUN ID gets assigned to the host).

ESX 5.1 and having Windows/Soliaru/Linix VM. they have 8 ESX 5.1 in a cluster, created Cascaded group one for shared storage and 8 MV for each ESX with the -Consistency_LUN option. Do we need to set boot lun id as 0?. Is there any specific settings for the HBA.

are you using Cisco UCS ? If yes, then boot parameter is configured in the boot profile. If another platform then you need to look at HBA settings. Typically yes, you want to use LUN ID 0 as your boot device.

Brocade DCX Switches. The customer would like to test by booting with the boot Lun id on one of the ESX server by  pulling out of ESX Cluster and then scan the common shared LUN's before proceeding with the cutover..Does this change modifies the copied data and make inconsistent. do i need to terminate and have another fresh set of OR copy before the cutover.

Please provide the cutover steps in a sequence. Thanks

i have created the OR Push using  differential option and will allow me to recreate and activate.Does testing the shared devices on the ESX server modifies the target data? Do i need to take another full copy by terminating the OR session or just recreate and activate does during cutover.

yes, you can activate and re-create but that will only copy changes that were made to control device after activation. OR does not know if there was anything written to target device. In your case you are going to start ESX server on target device, you are essentially modifying data. I would terminate the session after you complete your test and re-copy the whole thing from scratch. Hopefully your boot LUNs are small, ESX footprint is tiny.

i can't provide you step by step instructions as i don't know your environment. I don't believe you will be able to re-create OR session after you activate the session and write to target device. SDDF does not track changes on target devices, control device only (Open Replicator, i know it's different for TimeFinder/SRDF)

Dynamox is correct,  OR push does not keep track of changes on the remote device, if there was any writes to the remote device, you will end up with corruption. For the cutover steps, i cannot provide you specific details as I am not sure which servers you are using, for Cisco UCS it is much easier than HP, IBM or DELL but anyway, you should follow this:

1- Activate your first cycle of OR-PUSH copy for your esxi Boot LUNs. (Kindly note that you should use OR for boot LUNs only not your datastore.For datastores , Storage vMotion will be the best option)

2- Keep your data incrementally synced on a day to day basis. for instance if your first cycle was run on Monday and the planned cutover is on Thursday then run incremental copy on Tuesday and Wednesday to minimize the no of  modified tracks to be copied to the remote device in the cutover - (In your case it should not be too much data as boot partition for Esxi usually in terms of 5 GBs)

3- Shutdown your ESXi hosts group by group (Enter maintenance mode then shutdown) in the same cluster and run vMotion for the powered on VMs (DRS should take care of this if it is enabled) So, if you have 10 Esxi hosts in a single cluster then work on 2 hosts or 5 hosts at a time to eliminate any downtime.

4- Run one more incremental copy before the cutover and after shutting down your hosts.

5- Change the zoning and configure the boot LUN ID either form the Bios or from the server manager (For Cisco UCS this should be configured from UCSM)

General info:

• For the boot LUN ID, ESXi does not need to have LUN ID 0 to boot, For instance, in vBlock system where we have boot from SAN configured in Cisco UCS, We usually have a Masking view created with a storage group contains all boot luns for a specific cluster in your environment with LUNs have different LUN IDs

• Multipathing is not a big issue for ESXi (It should work fine) but for Windows, It is should be configured with a single path (Again this is for Cisco UCS and have been outlined in their docs)

Hope that helps.

Mohammed Salem

M.Salem wrote: Multipathing is not a big issue for ESXi (It should work fine) but for Windows, It is should be configured with a single path (Again this is for Cisco UCS and have been outlined in their docs) Hope that helps. Mohammed Salem

what do you mean Windows should be configured with a single path ? We are not building from scratch here, we are cutting over to another array, all paths should be presented at cut-over time.

Correct Dynamox, That is why I mentioned this under General info not part of the cut-over steps.

Hope that helps

Mohammed Salem

10-4

you can add Linux to that list as well ..Anakonda is not very happy when it sees multiple paths (at least in my experience with RHEL 5)