Hi,
I have several linux servers an also some windows vm in a 3.1 cluster (no ha): of the windows ones, some active directory controllers, some serving shares. Some windows share are also mounted atm by some linux (server) guests, and I am starting to use AD authentication on some linux servers, too. All important vm/ct are on shared storages (two nas boxes in the lan). I am wondering how to sort out the best startup/shutdown plan for my pve clsuter cts/vms.
Up to now I basically not planned anything (cold boot of all nodes is not a frequent event, obvioulsy, usually happens only when there's a planned/unplanned power failure beyond ups batteries capabilities..), and I manually fix most problems by hand on dependant servers, if something happens, (I work in the same building where as all nodes are, it's really the other room, so nothing difficult to do), but it's been a while since I started thinking that this should be managed better with pve vm/ct startup options.
Since the shared storages, there's the need to wait for their LUNs/volume to be available via iSCSI/NFS. Then, since windows servers are (usually) a bit slower to start (I have 2003/2008r2 versions), it would probably be much better to start those windows server before others, so their services should be available before other dependant servers boot up (ie: mounting win shares in linux could cause troubles, if either the share server or the auth server are not available...).
I've seen http://pve.proxmox.com/wiki/Virtual_Machine_Startup_and_Shutdown_Behavior, and read that <<Boot order is per-host and not per-cluster feature.>> so I am guessing how to deal with all machines starting on a node, and dependant other machines starting on another.
Just, before start experimenting on my own, I would like to hear if anyone hase good advices to share.
Thanks,
Marco
I have several linux servers an also some windows vm in a 3.1 cluster (no ha): of the windows ones, some active directory controllers, some serving shares. Some windows share are also mounted atm by some linux (server) guests, and I am starting to use AD authentication on some linux servers, too. All important vm/ct are on shared storages (two nas boxes in the lan). I am wondering how to sort out the best startup/shutdown plan for my pve clsuter cts/vms.
Up to now I basically not planned anything (cold boot of all nodes is not a frequent event, obvioulsy, usually happens only when there's a planned/unplanned power failure beyond ups batteries capabilities..), and I manually fix most problems by hand on dependant servers, if something happens, (I work in the same building where as all nodes are, it's really the other room, so nothing difficult to do), but it's been a while since I started thinking that this should be managed better with pve vm/ct startup options.
Since the shared storages, there's the need to wait for their LUNs/volume to be available via iSCSI/NFS. Then, since windows servers are (usually) a bit slower to start (I have 2003/2008r2 versions), it would probably be much better to start those windows server before others, so their services should be available before other dependant servers boot up (ie: mounting win shares in linux could cause troubles, if either the share server or the auth server are not available...).
I've seen http://pve.proxmox.com/wiki/Virtual_Machine_Startup_and_Shutdown_Behavior, and read that <<Boot order is per-host and not per-cluster feature.>> so I am guessing how to deal with all machines starting on a node, and dependant other machines starting on another.
Just, before start experimenting on my own, I would like to hear if anyone hase good advices to share.
Thanks,
Marco