Array activity causes SMB to become unresponsive (Cache Mover or SMB File copy/moving)

[Reiject]

Hi, I have been struggling with a problem since moving back to UnRAID (I had taken some time running my NAS on FreeNAS for a couple of years) It seems that as soon as there is any file/disk activity on the array (cache excluded) the system becomes incredibly unresponsive, to the point where it affects Windows machines which have mapped network drives mapped to it.

If the share is mapped \\hostname it is especially bad, using the IP address for the drive mapping seems somewhat more reliable.

I run a Plex server on a separate Windows computer (not a docker or VM) and if any array activity is occuring (be that the Mover or copying/moving files around) then the plex streams will start buffering or fail entirely, citing the file as being unavailable. Windows sometimes marks the mapped drives with an X until the activity is ceased.

When I used unRAID in the past I never encountered this, even when running the mover while multiple Plex streams were active.

Using the Netdata docker, I can see that IO-WAIT hogs a lot of CPU time. and any activity seems to cause really high "disk backlog"

 

PLease could someone who understands the background behaviour of the system help me in identifying the cause of this? I feel confident is not normal as when I used to run unraid on an i7 930 X58 based system I never had any of these problems, however that was with only 4 or 5 drives back then and now I am running 7 (9 including cache and parity)

Current system is a HP Proliant ML10 Gen9 with XEON E3-1245 v5 (upgraded recently from a G4400) with 64GB RAM and running a mix of WD Red 6TB drives alongside a couple of WD Black 4TB drives and Samsung 860 Pro 512GB for the cache

 

 

reijectserv-diagnostics-20201221-1333.zip

[Reiject]

bump - ongoing issue i am struggling to articulate to find other reports of similar issues

 

any help appreciated

[JorgeB]

This issue, having very high system load when using the user shares it's not new, but it only affects some users (or some hardware) and it's difficult to reproduce, to minimize it you can try running the mover only during off hours and/or using disks shares whenever possible.

[FixYouDeveloper]

@Reiject did you ever manage to fix this issue? I'm having similar problems - would love to know if it'll be better to move to freenas instead / buy a synology nas.

[Reiject]

Sadly I never resolved it. I think it is something to do with the hardware (but not sure exactly what) as I have also run TrueNas (FreeNas) on the same hardware and had the exact same issue (with transfer speeds dipping to 0 intermittently, sometimes causing windows SMB activity to error/fail)

My long term plan is to replace the hardware (new motherboard, processor etc.) but have not gotten around to doing it yet.

What hardware/specs are you using out of curiosity?

[sjrahn]

I am encountering this problem and I feel like it's only just started happening today.

 

The SMB share from my unraid box is mounted to a linux machine connected over a 2.5GBe link. As soon as I invoke the mover it totally distrupts that mount. Doing a simple list just hangs indefinitely. If I stop the mover and wait ~5 minutes it becomes responsive again.

 

I just switched the mount to nfs4 (which I wanted to do anyway) and I am not experiencing these same issues.

 

EDIT: Nevermind it's happening with the NFS shares as well. A little while after turning off the mover it magically comes back to life.

Aug 12 17:26:56 sjr-plex kernel: nfs: server ecba.local OK
Aug 12 17:26:56 sjr-plex kernel: nfs: server ecba.local OK
Aug 12 17:26:41 sjr-plex kernel: nfs: server ecba.local not responding, still trying
Aug 12 17:26:41 sjr-plex kernel: nfs: server ecba.local not responding, still trying
Aug 12 17:25:57 sjr-plex kernel: nfs: server ecba.local not responding, still trying
Aug 12 17:25:33 sjr-plex kernel: nfs: server ecba.local not responding, still trying
Aug 12 17:21:41 sjr-plex kernel: nfs: server ecba.local not responding, still trying

 

Edited August 12, 2023 by sjrahn