Last night (gmt +0) I switched to a new homelab server that I have have finally finished putting together. I installed Proxmox on the new one and moved everything over from the old server, that also ran Proxmox.
I had a couple of SSDs in the older server with an installation of Windows Server 16 Essentials on it, it did run a bit sluggish - I thought that was just down to the old server being the old server.
When I transferred the SSDs to the new Proxmox server and got the Windows Server back on it, it was still sluggish. - I pulled the drive and tested on CrystalDiskInfo and CrystalDiskMark; CDInfo said the drive was good, although in CDMark I was getting read speeds of 540MB/s and write speeds of 60MB/s - This was the case for 2 out of the 3 SSDs I had in the old Proxmox server.
Is there some SSD eating monster in Proxmox that kills SSDs with a default configuration install of Proxmox?
The new Proxmox server is on the newest version of Proxmox with 2 brand new SSDs being utilised (utilised is not spelled with a z, spellcheck -_-)
Is there anything I can change, and or do to stop anything on Proxmox potentially killing my new SSDs?
( I am not using enterprise level SSDs, just off the shelf consumer grade ones )
TL;DR version: Does Proxmox kill SSDs? - How do I stop it if it does?
Sorry for the late response; I made a similar post over at the Proxmox user forums. Some of the responses are as follows:
Proxmox does not “kill” SSDs.
SSDs has limited write cycles, so you need to make sure that your SSD are suitable to your use case (write workload)
First, you should look on the write endurance of your SSD model. - tom, May 14, 2018
On the other hand: PVE will write a lot of data even in a homelab and may kill your SSD faster. This is still true.
Using a HDD for the PVE installation and consumer SSDs as Datastores is my “solution” to that… minor inconvinence. - tschanness, May 14, 2018
One way to get more life out of the SSD drives is to over provision them. I have pretty cheap SSD with 256GB but actual partition size is 100GB with rest as unallocated. That will help with wear leveling. - Tahsin, May 15, 2018
if you don’t have a cluster there are some services you can turn off which continuously write to the disk, thereby extending life significantly.
pve-ha-crm and pve-ha-lrm specifically are the ones to target - frenchfries, May 16, 2018
I can confirm. Installing on SSD will kill it if you not stop and disable the 2 pve-ha services. The solution is installing over on a mirror of little mechanical disks. It’s a pity because in this way you waste 2 disk bays.
It could be a good way to write the cluster status on ram on every node. - tuonoazzurro, May 24, 2018
