Full flash PVE

Alessandro 123

Well-Known Member
May 22, 2016
653
24
58
41
Anyone using PVE on a full SSD server?
Any troubles?

I'recently read that even enterprise SSD tend to fail suddenly without warnings and much faster than the expected (like a Samsung SM863 failed after about 8000 hours, less than a year)
 
Write pattern on ZIL is different from the pool ?

Anyway, i wasnt aware of SV843 model.
Any difference from SM893?
 
Write pattern on ZIL is different from the pool ?

I have no external ZIL, everything on the 6 disks in 3 mirrored vdevs and smartctl reports everything healthy at 99% and Proxmox VE reports 1% wearout. We have this disks for quite long.

With the smartctl TUI from http://www.jdgleaver.co.uk/blog/2014/05/23/samsung_ssds_reading_total_bytes_written_under_linux.html

Code:
root@proxmox4 ~ > ./samsung_ssd_get_lifetime_writes.bash
------------------------------
 SSD Status:   /dev/sda
------------------------------
 On time:      20,795 hr
------------------------------
 Data written:
           MB: 13,260,996.935
           GB: 12,950.192
           TB: 12.646
------------------------------
 Mean write rate:
        MB/hr: 637.701
------------------------------
 Drive health: 99 %
------------------------------

Anyway, i wasnt aware of SV843 model. Any difference from SM893?

I do not know. We purchased the disks from the hardware vendor that sold us that particular server due to warranty reasons.
 
There is a post on ZOL ML talking about suddenly failure , with no warnings, at about 8000 hours and 29TB written

A reply is saying that the same happened with Intel SSD at about the same number of hours
 
I think SSD Failure has nothing to do with proxmox, I am using non-enterprise grade SSD (cheap one) since 3 years for Proxmox, still working great.
 
Up to now, we implemented almost over 100 SSDs in different setups mostly on Fujitsu hardware without any disk failure (all non-PVE, pure Oracle databases). Up to now, everything is working great.