Napp-it - NFS Read bottleneck?

NOTORIOUS VR

Member
Nov 24, 2015
78
7
8
41
I've noticed lately that once there is some load on the server, there seems to be a heavy (random?) read bottleneck that I really do not know how to put my finger on.

dd writes seem to be ok, but reads either get stuck or something - not quite sure at this point.

eg. around 3-4 Plex streams going at once, I will have troubles (on my LAN) starting a new stream. Regardless of device, regardless of connection (WLAN, LAN 1G or 10G).

SSH into the server, everything seems fine until I try to ls/ll the folder, then there is just a delay even to list the dir.

time dd TO the NFS share which has the video data works ok it seems with 204 MB/s:
Code:
sascha@mediasrv1:/shares/video$ time dd if=/dev/zero of=/shares/video/test2.img bs=16k count=128k
131072+0 records in
131072+0 records out
2147483648 bytes (2.1 GB, 2.0 GiB) copied, 10.5507 s, 204 MB/s
time dd FROM the NFS share just get's stuck because it's absurdly slow and I don't understand why:
Code:
sascha@mediasrv1:/shares/video$ time dd if=/shares/video/test2.img of=/dev/null bs=16k
^C
^C^C^C34581+0 records in
34580+0 records out
566558720 bytes (567 MB, 540 MiB) copied, 358.44 s, 1.6 MB/s

real    5m58.443s
user    0m0.016s
sys     0m0.420s
Reading from the same share via my Windows desktop connected via 10gbe seems to be ok as well:
1.jpg

So it seems there is something going on between that linux server and my storage server - not sure what tools to use to track this down.

both VM's are on the same host, both use a dedicated vswitch/nic/subnet for storage.

napp-it vm is 4vcpu - 40GB RAM
the media share is Z2 with 10 spinners in one vdev.


Anyone who has a suggestion - I would really appreciate!

Thanks
 

gea

Well-Known Member
Dec 31, 2010
2,870
1,013
113
DE
i would
- check local performance (test sequence at pool > benchmark
- check network (iperf)
- optimize nfs and ip settings system > tuning)