DRBD primary / secondary

primary/primary is the tested and working configuration for doing live migrations - why does a split brain scares you?
 
tom said:
why does a split brain scares you?
Click to expand...

In the article http://pve.proxmox.com/wiki/DRBD I read, on the "Recovery from communication failure": "Normally one would choose to discard the data of the least-active node".
I think if I use the primary/secondary instead of primary/primary, this does not happen..... I am no expert ;-)

What do not I understand is this: can I use drbd in primary / secondary mode and migrate virtual machines from one server to another?

eg.
- 2 proxmox on cluster (proxmox-1 and proxmox-2)
- 2 disk for every proxmox
- 1 disk for OS
- 1 disk for DRBD
- DRBD configured in primary/secondary mode (primary on proxmox-1 and secondary on proxmox-2)
- 3 KVM virtual machine on proxmox-1
- 2 KVM virtual machine on proxmox-2

can the 2 KVM virtual machine use "drbdvg" (see: http://pve.proxmox.com/wiki/DRBD) for store the virtual disk ?

thank you.
 
draguz said:
In the article http://pve.proxmox.com/wiki/DRBD I read, on the "Recovery from communication failure": "Normally one would choose to discard the data of the least-active node".
I think if I use the primary/secondary instead of primary/primary, this does not happen..... I am no expert ;-)

What do not I understand is this: can I use drbd in primary / secondary mode and migrate virtual machines from one server to another?
Click to expand...

no live migration, this works online in primary/primary.

btw, if you have detailed questions about DRBD, I suggest you post on the DRBD user mailing list.
 
tom said:
no live migration, this works online in primary/primary.

btw, if you have detailed questions about DRBD, I suggest you post on the DRBD user mailing list.
Click to expand...

OK, I understand.
Only one last question: an OFF-LINE migration is possible in a primary/secondary configuration ?

Thanks patience !
 
Only primary/primary is tested as this is the working solutions - if you choose any other setup you need to think by yourself, but why would someone for the second best solution if you can get full functionality?
 
tom said:
... but why would someone for the second best solution if you can get full functionality?
Click to expand...
Because I prefer migrate the virtual machine in off-line mode and don't risk to loose the data of the least-active node (in case of problem).
I'll do some testing with a primary/secondary DRBD configuration and post my result.

Thank's for your help.
 
I read more documentation on drbd and lvm.
Is impossible use primary/secondary configuration as a good solution because only one node have a full access to the phisical volume.
Now is clear, thank you again.
 
You can create two drbd resources - on the first you could store the VM's for the first node and on the second - the VM's on the second node. That way when split brain occurs, you can easily discard data - for the first drbd - discard data of the second node, for the second drbd - discard data of the first node.
 
fluke said:
You can create two drbd resources - on the first you could store the VM's for the first node and on the second - the VM's on the second node. That way when split brain occurs, you can easily discard data - for the first drbd - discard data of the second node, for the second drbd - discard data of the first node.
Click to expand...

great idea, thank you.
 
You must log in or register to reply here.
Top Bottom