wearout indicator

[gijsbert]

We are running proxmox on intel ssd's for over a year now. Although there is not very much disk activity, the wearout indicator is still 0%, even after using it for about a year now and I'm wondering if this is correct? Attached are two screenshots of what we see in the GUI of proxmox. Can anyone confirm the wearout indiciator is indeed 0% using Intel SSD's or does Intel have another smart ID when it comes down to wearout indicator?

Thanks in advance for any reply,

Gijsbert

 

[dcsapak]

we get the information directly from the ssd via smartcl
so unless intel changed the field number for their ssds, the 'media wearout indicator' show 0% wearout

 

[tom]

In one of our clusters, we also use Intel DC 3520 - I also thought there is an issue but now I see a 1 % wear-out in one of my OSDs. See screenshot.

 

[tom]

Just to add, this Ceph cluster is running for about a year, but we do not write that much on this one.

 

[drlight]

Hi. I've just paid attention that wearout indicator shows incorrect values. When smartctl was shown 10% indicator in pve was shown 0%. Now it's 1% while the real is 11%. See attached pic.
PVE Manager Version is 5.4-15/d0ec33c6

 

[dcsapak]

Now it's 1% while the real is 11%. See attached pic.
PVE Manager Version is 5.4-15/d0ec33c6

this is fixed since pve 6.1, so i'd recommend upgrading to a current (supported) pve version

 

[Wirehead]

Just an off-topic reply..
Since moving from ESXi to Proxmox, these wear indicators are -barely- moving for me. And I had these NVME drives in a mirror in an ESXi VM (via passthrough) presented back to the host. When moving to proxmox, I just re-imported the pool. And started using them. In a year, my drives went from 0 to +- 20% wear with hardly any use.

Now in 2 months of proxmox use, they've only gone up 1%. Extrapolate that data; and for 1 year of proxmox use: that's 6% (at worst). With ESXi and the same loads, I was at 20% somehow. That means the lifetime of my NVME drives has been extended by at least 3 times

As I log all this smart data to Grafana, I can quickly see trends. Once I moved from ESXi to Proxmox; I noticed that the temperature of my NVME drives was elevated the first couple of weeks, in several "batches" it seems. I suppose this was internal garbage collection in the firmware that was somehow unable to run in ESXi, as ZFS stats didn't show any I/O; but temperature doesn't lie . After those days, all seems normalized.

Bottom line: Hurray for Proxmox

 

[H4R0]

Your drives must be in the smartmoontools db https://raw.githubusercontent.com/mirror/smartmontools/master/drivedb.h

And it must be up to date on the server if its not the proxmox gui will always show 0% as default for SSD's.

It would be great if proxmox fails to retrieve the wear indicator to show N/A instead of 0%...