Proxmox VE + DRBD HA Storage

[dietmar]

pacemaker combined with openais also locks promising

http://clusterlabs.org/wiki/Main_Page
http://openais.org/doku.php

 

[turboloader]

For questions pm

 

[turboloader]

For questions PM

 

[dietmar]

How do you 'fence' the other system?

 

[dietmar]

But how can I benchmerk my filesystem ?

Benchmarking is difficult and time consuming ...

 

[turboloader]

For questions pm

 

[dietmar]

I think I will go back to my first configuration. I tried now to work with ocfs2 on a lvm of two harddisks per node in a primary primary cluster the performance of the vm's is very good. But I got problems with live migration, because both systems mounted the same drbd storage. So if I migrate one machine to the other it seems to work but then after shutting down my vm after migrating , my files of the vm disappeared.

Current migration scripts delete the original data after rsync. Of course, this is not what you want if you have shared storage ;-)

But vzmigrate has an option keep data:

vzmigrate --remove-area no [--online] [-v] destination_address CTID

- Dietmar

 

[turboloader]

For questions pm

 

[dietmar]

no, but i guess you can modify qmigrate yourself (just for testing).

the lines:

eval { run_command (2, "qm destroy $vmid"); };
logmsg (0, "destroy local vm failed") if $@;

needs to be disabled.

 

[turboloader]

For questions pm

 

[tecto]

active/active clustering can easily be overcommitted

i found the following discussion and howto to be very helpful:

http://forum.openvz.org/index.php?t=msg&goto=5775&
http://wiki.openvz.org/HA_cluster_with_DRBD_and_Heartbeat

i have an ubuntu cluster (heartbeat/drbd8) running openvz, would love to see this completely integrated into proxmox. there would probably need to be some resource reservation to prevent OOM issues in a node failure scenario.