Migrate HA from 3.4 to 4.0

[romapao]

hi, I have a datacenter consists of n. 2 servers of R620 + a nas qunap
I built a HA cluster with DRBD Proxmox VE 3.4 on both servers and using a quorum of iscsi LUN created ad hoc on Qnap and fancing with DRAC DELL
can I migrate on PVE 4? with DRBD9 and Fencing with hardware watchdog?
I remind you that both servers have a subsciption active.
thank you

 

[tom]

qdisk is not available in 4.0, so you need 3 nodes for HA. also, there is no in-place upgrade from DRBD8 to DRBD9.

so you need to redesign your setup, and re-install.

 

[romapao]

the third node must be identical to the other two in terms of hardware CPU and RAM DISK?

 

[tom]

the third node must be identical to the other two in terms of hardware CPU and RAM DISK?

No. you an use two big boxes and a small one just for the quorum.

E.g. A small Intel Atom based server.

 

[romapao]

OK thanks .
then I should
-perform Backup of VM,
-reinstall PVE 4 on the two big servers and on the small,
-build the HA cluster of DRBD9
-and finally restore the VM?

 

[tom]

basically yes. please note, DRBD9 is technology preview and not yet totally supported, still some missing features. DRBD9 is still very new, the drbdmanage is also not yet stable and feature complete.

 

[mir]

Just wondering. Since Debian supports ARM could a third node be running Rasberry Pi or other ARM based platforms supported by Debian?

 

[tom]

Just wondering. Since Debian supports ARM could a third node be running Rasberry Pi or other ARM based platforms supported by Debian?

No, Proxmox VE does not run on ARM.

 

[mir]

No, Proxmox VE does not run on ARM.

The stuff coded in C I guess (pmxcfs etc)?

 

[dietmar]

The stuff coded in C I guess (pmxcfs etc)?

yes

 

[Manta]

No. you an use two big boxes and a small one just for the quorum.

E.g. A small Intel Atom based server.

How do you make sure that the small Atom based Server does not take VM's over if one of the big Nodes dies?

 

[tom]

How do you make sure that the small Atom based Server does not take VM's over if one of the big Nodes dies?

With HA Groups, see:

http://pve.proxmox.com/wiki/High_Availability_Cluster_4.x#HA_Groups

 

[Manta]

Thank you! Cool Feature!