One PVE node (no cluster) with Nimble over FC

O

ocampofer

New Member
Aug 20, 2024
4
0
1
Good morning.
I have a DL360G11 server with FC boards
I also have a Nimble with FC and 2 controllers.
I'm not going to form a cluster in Proxmox, I'm just going to use a node called SRVPVE01.
Can I present the LUN to the SRVPVE01 node, configure the multipath, and on that LUN, create a LVM-Thin, and use the disks of the VMs and Containers in RAW format to have snapshots and make better use of the space?
Is this recommended or is there a better way?
I can only do it through FC.
I already have this hardware, and I want to do the best I can with it.
Greetings and thanks!
 
Using 2 lun's, one from each controller and 2 times that multipathings, would give you more I/O oompf if you have enough disk for.
If 1 controller fail both lun's are still served be the remaining.
What are your FC connections, 4x FC32 ? What are your disks and setup ?
 
bbgeek17 said:
Hi @ocampofer , welcome to the forum.

Yes, you can do what you have described as long you only have one server connected to the LUN. And, yes, its a good strategy.

Good luck

Blockbridge : Ultra low latency all-NVME shared storage for Proxmox - https://www.blockbridge.com/proxmox
Click to expand...
Thanks for the reply, I think that's the path I'm going to follow.
waltar said:
Using 2 lun's, one from each controller and 2 times that multipathings, would give you more I/O oompf if you have enough disk for.
If 1 controller fail both lun's are still served be the remaining.
What are your FC connections, 4x FC32 ? What are your disks and setup ?
Click to expand...
I'm going to do some research on what you're telling me.
I always use 1 LUN with all the available space.
It's an old Nimble, a CS1000 with 11 1TB HDDs and 2 1TB SSDs of cache.
It's 2xFC16GB, one for each controller, and they are active/passive.
That's why I was thinking of 1 LUN
Thanks for your comment, I'm going to do some research and see if I can get 2 more FCs to test.

The DL360 has four 480GB SSDs, forming two RAID1s with two disks.
One of the RAID1s now contains the PVE installation and the other RAID1 contains an LVM-Thin where I have some test VMs hosted, and where I will temporarily place a DC and a DNS, so as not to leave the company without a system for a long time during the migration.
I was thinking of doing the migration with Veeam, using the current backups of all the VMs and having it take care of the conversion.
Since currently, the VMs are on a DL360 Gen10, using that Nimble, and I have to take out the FC cards to put them in the Gen11.
 
Last edited:
I just remembered, I have a brief question.
In ESXI, the storage took longer to turn on than the server, and it caused me problems.
With PVE, can I delay the powering on of the VMs until the Storage is ready?
The DL360 only lets me delay the powering on for 120 seconds, and at least with ESXI that wasn't enough.
Regards and thanks!
 
ocampofer said:
I just remembered, I have a brief question.
In ESXI, the storage took longer to turn on than the server, and it caused me problems.
With PVE, can I delay the powering on of the VMs until the Storage is ready?
The DL360 only lets me delay the powering on for 120 seconds, and at least with ESXI that wasn't enough.
Regards and thanks!
Click to expand...
Although you can put a delay on VM start : https://forum.proxmox.com/threads/delay-start-of-a-vm.111212/
If your storage takes too long to physically start, you may still have issues. The only way to know for sure is to try it.

Blockbridge : Ultra low latency all-NVME shared storage for Proxmox - https://www.blockbridge.com/proxmox
 
ocampofer said:
I always use 1 LUN with all the available space.
It's an old Nimble, a CS1000 with 11 1TB HDDs and 2 1TB SSDs of cache.
It's 2xFC16GB, one for each controller, and they are active/passive.
That's why I was thinking of 1 LUN
Thanks for your comment, I'm going to do some research and see if I can get 2 more FCs to test.
Click to expand...
Then you are right with just 11 hdd in maybe raid7 (??) you don't have enough disk for raidset of other controller and 2 volumes on same raidset isn't good for performance, so 1 LUN is right.
 
waltar said:
Then you are right with just 11 hdd in maybe raid7 (??) you don't have enough disk for raidset of other controller and 2 volumes on same raidset isn't good for performance, so 1 LUN is right.
Click to expand...
I really appreciate your help.
Thanks in advance.
One last question, the server has 2x4510 Xeon Silver, and 384GB of ram.
For example, for an 8-core VM, should I use "host" 2 sockets, 4 cores and NUMA enabled? or 1 socket, 8 cores and no NUMA?
 
If you want to give a vm 8 cores I would do it as 1socket with 8cores, no numa. And be aware don't give a vm to much memory as it's much better to have the mem on host instead of blocked virtual in vm, check mem usage inside for swapping to minimize mem definition.
 
You must log in or register to reply here.
Top Bottom