Running 2 VM with same disk on DRDB

[ripp2003]

Hi all

Here the config

· 2 Servers installed with Proxmox 4.4 as SA and SB

· On each server, there are 2 logical partition (20Go as PA and 500Go as PB)

· Proxmox is installed on PA

· With the tuto from the wiki, I’ve installed DRDB9 on PB

· The 2 servers are in a cluster

Everything is OK, I’ve build a VM with Windows 7 image on DRDB on SA and the image is well replicated on SB.

I can move my VM between the servers so Cluster is OK

BUT I don’t think I've well understand….

When I’m running the VM on SA, and I shutdown SA, my Windows ends…..so I can't migrate anything...


I thought SB will take relay because the image is present on SB (with DRDB) but no…..


To do this, I’ve made a trick and even if it works, can you tell me if it’s OK because Proxmox can’t do this natively, I’ve to edit conf file:

· I’ve made an identical VM config on SB

· I’ve copy the content of 100.conf on SA to 101.conf on SB


So now, each VM use the same image disk (wich is in DRDB) and if the SA if off, I can use the VM on SB with same data (because of DRDB)
My goal is to have always a server wich can run a VM with same data (and I'don't have any SAN or network storage)

Is it good ?

THanks

 

[Xahid]

Have you created HA Group and include the VM in that group ?

 

[ripp2003]

Thanks for this quick reply : no, I don't have done this.... I will test this Monday and report here.
Really thank you for helping me ;-)

 

[Xahid]

before you do this, I suggest to revert back all the changes you made in 100.conf files.

 

[ripp2003]

Hi sorry for this very late answer......
Got lot of work with Proxmox
So, I've tested what'you 've said and, it's not what we want.

We finally use another solution because after some test, DRDB is not so strong and we faced some errors.

So now, we have 1 linux on A, 1 linux (a clone) on B and synchronisation is made with "built in" synchro of mysql.

Ou company accept some delay when a server fail...

Thank you again ;-)

 

[Xahid]

What went wrong?
Your other solution is also good as long as it serve the purpose.

You can also try two nodes drbd84 cluster with heartbeat (for HA), drbd will take care of sync. while heartbeat will take care of Services, In this scenario you don't even need Proxmox.

 

[ripp2003]

DRDB fail with "pending operation" because I've delete a file with one drdb host offline. Never could synchronize again...
We need simple storage (and drdb add new layer) with really simple administration.
HA can't be used because we need a quorum...and we only have 2 servers.
For now, I've got to set "pvecm e 1" so when one server is offline, the other can start VM