Super low FSYNCS/SECOND and I/O delay

L.O.S

Member
Apr 7, 2021
30
1
13
43
Hello All,

I am real new to Proxmox and I am find some odd behavior that I can't explain:
I am running the system on an MVME drive and have 2 pools - first: 6x 240G SSD (ZFS RAID10) and the second 2x 1T SSD (ZFS Mirror)

the rest of the server is :
2x XEON E5-2680v4
128G ECC Mem

the pveperf gives me:

CPU BOGOMIPS: 268919.28 REGEX/SECOND: 3175928 HD SIZE: 56.84 GB (/dev/mapper/pve-root) BUFFERED READS: 1569.19 MB/sec AVERAGE SEEK TIME: 0.05 ms FSYNCS/SECOND: 310.44 DNS EXT: 23.97 ms DNS INT: 6.96 ms (home)

first: 6x 240G SSD (ZFS RAID10):
CPU BOGOMIPS: 268919.28 REGEX/SECOND: 3325210 HD SIZE: 55.18 GB (StoreZilla) FSYNCS/SECOND: 14.88 DNS EXT: 16.25 ms DNS INT: 7.23 ms (home)

second 2x 1T SSD (ZFS Mirror)
CPU BOGOMIPS: 268919.28 REGEX/SECOND: 3324830 HD SIZE: 826.81 GB (FileZilla) FSYNCS/SECOND: 838.28 DNS EXT: 16.42 ms DNS INT: 7.62 ms (home)

I am running all from the MB with no raid or anything fancy.

I am pondering going for some Intel D3-S4510 960G - would you consider that a good choice ?


Tanks for you help
 
Last edited:
Indeed, they are really shitty for zfs/ceph :p
the 2x 1T are 870 EVO (still pretty bad from the stats) and I can see notable improvement.

I ordered a couple of DC S4610 - I hope to see real improvement :fingers_crossed
 
ashift=12?
I am guessing not ?

Code:
Disk model: CT240BX500SSD1 
Units: sectors of 1 * 512 = 512 bytes
Sector size (logical/physical): 512 bytes / 512 bytes
I/O size (minimum/optimal): 512 bytes / 512 bytes
 
the output is super long, see attached.

Cheers
conf looks ok.
it can be due to ssd (bx500) cache full. Anyway, do not use bx500 for write big data sizes (tens Gigs).
You can use not expensive (for example) kingston dc500m.