Adding additional HDD to Proxmox Server

[Chiranjeevi]

Hi

Our Server has 4 no's of 10TB HDD and with RAID1 so 2 drives
My question is about creating the Storage for the VMs,
i have installed Proxmox in drive 1, which has local and local-lvm storage where local is given 100GB and 9.9TB for the local-lvm, where we keep VM's right now.

i need to add more disk space by using drive 2. So, which is better solution? extend the local-lvm or create another lvm with different name?

These are 2 things i can think, Can you help me what is best option

1: If Extending the local-lvm
a: What will happen if the Drive 2 Fails, how my VM's are separated?
b: I think this option is better when we do cluster, because same LVM is present in Other Server.

2: If we Create a Seperate LVM-Thin store using Drive 2
a: we store the VM's in different drive easy to maintain i think.
b: if cluster is configured, either we have to make same LVM in Other servers or we should not move VM from one server to other.

Thanks
Chiru

 

[udo]

Hi Chiru,
two raid-1 with 2*10TB hdds sounds not very fast. I assume you have an big IO-bottleneck.

I would use all four disks in raid10 to extend the IO.

To your storage question: It's equal if you use one lvm-storage or two - both are local storage and not shared storage - even if the storage has the same name on different nodes - it's local storage and the vm-images must migrated too, to run an VM on the other node (and if one node die, the VM-images are not accessible).

For easy migration you need shared/distributed storage, like nfs, DRBD, Ceph, lvm on FC/SAS/iSCSI-Storage.
Unfortunality the drbd-support isn't quite good these days...

And for ceph you need three nodes and with 1 10TB-disk on each host (or two), I assume you will not get much fun.

Perhaps zfs and zfs-sync for recovery if one node die?!

Udo

 

[Chiranjeevi]

Thanks for the reply, and sorry for late reply from me.

Yes you are right we are facing more IO-delay while we copy the VM or While taking backup, we are not able to use the other running VM's i have to change that to RAID10

But still one thing i need to clear is,

If i extend the size of the LVM (local-lvm) by adding another disk, how can i have the view of VMs are in which disk?
for example i have /dev/sda1 which is 10T and i created the local-lvm.
now i add another drive 10T which is /dev/sdb1 and extend the local-lvm.
and created 10 VMs
now how i come to know which VM is on /dev/sda1 and which VM is on /dev/sdb1?

Because my worry is if any disk one disk failed which all the VMs not going to run i need to know.

-Chiru

 

[guletz]

If i extend the size of the LVM (local-lvm) by adding another disk, how can i have the view of VMs are in which disk?

You can not know. This info it is at the lvm level. lvm know only about blocks of disk.

The best for IOPs is raid10. Take in account that your disks are very bigger. So it is possible that when one disk will fail, the time to reconstruction of the raid will be very long. In this time it is possible that another disk to fail.

If you have valuable / critical data, this could be a problem. From a balanced perspective, it will be a good solution to use zfs (aka raid10). One big advantage will be that you will have 2 advantage compared with lvm(2 disk like raid0):
- you can use pve-zsync as @udo say
- block compression (this will virtualy increase your IOPs)

But for zfs you will need RAM only for this. At minimum you will need 1-2 Gb, depending on your usage case. More memory will give you better results, but depends .... how is your case.