Single node with 2TB SSD with ext4 will it wear out too soon?
- Thread starter logui
- Start date
Not sure in which context you are asking, i.e. considering other hardware or other filesystem. The Samsung has 1200 TBW, some people would recommend PLP drives (albeit with lower TBW).
Ext4 (or XFS) will definitely spare you of extra writes, if you are concerned about excessive writes, you have to consider what you run (e.g. OPNsense and configure it to avoid them, e.g. [1]). You should probably run it first and see what is eating your TBW, then tune, keep checking SMART values.
Also consider firmware updates for the SSD, as there were some well publicized hiccups with e.g. 980 Pros I am aware of, but I am not keeping up with the recent news on those.
[1] https://docs.opnsense.org/manual/settingsmenu.html#logging
Ext4 (or XFS) will definitely spare you of extra writes, if you are concerned about excessive writes, you have to consider what you run (e.g. OPNsense and configure it to avoid them, e.g. [1]). You should probably run it first and see what is eating your TBW, then tune, keep checking SMART values.
Also consider firmware updates for the SSD, as there were some well publicized hiccups with e.g. 980 Pros I am aware of, but I am not keeping up with the recent news on those.
[1] https://docs.opnsense.org/manual/settingsmenu.html#logging
Thank you, I have checked the FW and is up to date, is it safe to assume that the default installation of Proxmox with 1 node has Cluster Services disabled? I ask because I have read that Cluster Services perform a lot of writing to the disk, therefore wearing out sooner the SSD
Now this is a contentious topic nowadays, especially if you ask me, out of all people.
You can have a look at these links:
https://forum.proxmox.com/threads/etc-pve-pmxcfs-amplification-inefficiencies.154074/#post-701249
Yes, they are collected by me and counting. So no, you cannot assume that the "cluster services" are disabled. I am sure in some of the threads there's a remark on disabling (or better yet masking) HA services to avoid some of the writes. But you get the idea.
I also want to add that everyone (so far) in that thread is correct. But it's only you who can monitor what your clean install goes on with writing (and report back perhaps?).
fyi, here running two hosts 24/7/365 from an USB Key without problem yet , one since 2021 , the other from 2022.
https://forum.proxmox.com/threads/running-proxmox-from-a-usb-drive-solved.66762/post-464877
imo, only ZFS is the most TBW eater then require DataCenter/PLP disks.
https://forum.proxmox.com/threads/running-proxmox-from-a-usb-drive-solved.66762/post-464877
imo, only ZFS is the most TBW eater then require DataCenter/PLP disks.
I am all for fun, but I just cannot help and have to add a word of caution to the OP - deploy what you need and check on SMART, if you discover something amiss, you need to tweak the setup - this does not have to be ZFS, it does not even have to be pmxcfs. PVE is definitely not the kind of hypervisor you would run off an SD card without tweaks.Now that I have been doing all my funny tests with pmxcfs, I ended up running the whole thing off RAM-disk and I can definitely confirm there are situation where lots of writes can happen.
I am aware of all sorts of tricks like log2ram [1] and pmxcfs-ram [2], but I also like "professional" solutions, so I even left some comments for the authors (e.g. [3]) to get back to me if they wish.
Most people ending up on the forum have some combination of the factors adding on top of each other. But I also believe the old adage that on a long enough timeline the survival rate approaches zero.
So I would keep monitoring such setup for any anomalies. Surely not just place it on a USB flash drive and forget.
[1] https://github.com/azlux/log2ram
[2] https://github.com/isasmendiagus/pmxcfs-ram
[3] https://forum.proxmox.com/threads/o...am-based-storage-question.123674/#post-703044
Last edited: