Forum Discussion
So are you saying multipathing is not supported with VCS 6.0 agents. As the agent guide says you no longer need to use DiskReservation agent (like you had to in 5.1) and it gives example showing multipathing, this sort of implies it is supported, unless multipathing examples only applies to DMP, but the guide certainly does not say this.
I still don't understand why this has to be tested - let me draw an analogy. Supposing someone is using the Oracle agent and they install some other software and this causes Oracle not to be able to start, so they think the Oracle agent is broken as the VCS resource won't online. If they were to log a call, then my expectation is one of the first things that is asked (if the logs don't show what the issue is) is does it start manually outsdie of VCS control and if it doesn't then the issue is not VCS and the customer needs to resolve the issue. This is how I see the LVM agents - if some other software is installed (like third party multipathing) and this causes problems activating or deactivating (or importing and deporting) the Logical Volume Group, then this is not an issue with the agent - it is an issue between LVM and the third party multipathing softtware. The only exception is if using third party multipathing causes you to activate or deactivate the Logical Volume Group differently (maybe a different flag would be required, but this seems unlikely) or if the command the agent uses to monitor the Logical Volume Group has different output (with all disks paths up or one down) - and for this I can only see this possibly happening with O/S multipathing software because with hardware multipathing like EMC Powerpath, I believe the sinlge multipathed disk is added to the Logical Volume Group (not the separate paths) and so the disks looks no different to LVM than any other disk and if a path is lost, then LVM is unware.
Therefore unless, I am wrong and the LVM agents does some lower level stuff (agent docs for linux just say the vggimport, vgexport and vgdisplay commands are used), then I think all hardware multipathing should be supported and preferably O/S multipathing and I wouldn't have thought this requires much testing (mainly need to just make sure monitor entry point works if command shows differerent output if multiple paths are used, in particular if one is disconnected).
Mike
Related Content
- 11 years ago
- 11 years ago
- 2 years ago