Intermittent freezing on video playback

Status
Not open for further replies.
Joined
Jun 13, 2013
Messages
25
So just looking for some pointers on this.

I recently moved my video collection from an old pool over via snapshot copy to a new one on the same machine. Old pool was a raidz1 of slow 2TB drives that were aging and starting to fail. New pool is raidz3 on faster 3TB drives. Nice increase in performance and I sleep better thanks to decent redundancy.

However, when playing the files on my media players (sharing is via SMB), I am seeing frequent, though not predictable, stalls in playback. This evidences itself as buffering (but there's no read activity going on) which occasionally results in the media player totally stopping playback. I would very occasionally see this on the old pool, however nowhere near as frequent. Watching a movie now this happens at least once, whereas previously it might be one in ten, or less. Playback stats show the buffer dries up and there's no data coming in. This happens seemingly at random during periods of no other load (as far as I can see at this stage).

Things I've tried so far:
  • Copying files that aren't cached from the volume. Solid, consistent reads, verified by watching zpool iostat -v, reports tab in the GUI, network traffic etc. I am confident there isn't an issue here given we're talking about, on average 4-6GB files over a 2 hour period, and I can see the pool sustaining 100MB/s+ sequential reads when copying. Network (all gigE) seems to sustain this fine too, so I don't think it's a network issue either.
  • Different media players. Two separate linux based OpenElec builds, on two physically separate PC's on separate (wired) network runs, and an Android based Kodi install.
  • All new disks were burnt in before use (long, destructive smart tests) and still check out fine.
  • System generally seems ok. It's a dual quad core Xeon X5570 2.9Ghz box with 48GB ram that's very lightly loaded. Old, but it's not really doing anything difficult.
  • SMB shares are created the same as previously. They have aio_pthread and streams_xattr set - but I only changed the path to the point to the new datastore, so these were set previously.
  • * Pool is healthy, 53% used, no de-dupe, no compression, no encryption, no periodic (or any) snapshots, no L2ARC, ZIL etc. Bog standard. New pool was created via the GUI and the snapshots from the old pool copied over.

Things changed recently:
  • New pool
  • Updated to latest build on the 9.10.2 stable train (was on U5, now on U6)

At this stage I'm just looking for ideas on what could be happening and any suggestions on how to go about diagnosing and fixing?
 
Joined
Jun 13, 2013
Messages
25
Doing some searching on the forums here, I found a few people with similar symptoms in the past, with solutions ranging from esoteric samba settings to new hardware.

Going to try NFS instead of CIFS for the media boxes as a test later on to see if that makes any difference.
 
Joined
Jun 13, 2013
Messages
25
Tried NFS and watched a film last night and same symptoms. So it's not CIFS/Samba.

Nothing obviously wrong in most of the graphs, except for the pending I/O requests. These are (to my mind) odd, as they're showing 300m of pending i/o requests - however the total iops on the disks are in the 5-10 range for the same period. See attached - I've picked on a single disk and arranged the graphs up so their timings are in line. I've included NIC traffic to show data still seems to be moving.

Any ideas what's with this?

My next plan is to create a quick pool on the older disks, copy some files and do some tests with that.
 

Attachments

  • stall2150.png
    stall2150.png
    37.1 KB · Views: 283
Joined
Jan 25, 2013
Messages
4
Hi Kristan,

I see this thread is a little old, but I was wondering if you were able to solve your issue?

My system started doing the same thing recently whereas before it was fine with zero issues. I have a similar setup; wired gigabyte ethernet, kodi on a HTPC running windows 10 that plays back the content, occasional buffering/jamming, shared via CIFS/Samba/SMB, etc..

The only thing I've changed since the issues started was a failing drive. I replaced it with a new 6 TB WD RED drive, but there aren't any errors reported, pool status is healthy, when the buffering happens I've been watching the log and nothing appears (in /var/log/messages, or any of the samba logs)... also I glace over at the machine that has the files on them and don't see any major activity on the LED lights on the drive bays.

But I can't ignore that I did switch a drive - same as you - and before that everything was fine. Did you end up trying your old drive(s)?
 

Apollo

Wizard
Joined
Jun 13, 2013
Messages
1,458
You should try running Netdata that comes with Freenas 11 (it part of the Services in the GUI). I find it to provide very valuable data for realtime metrics.
 
Status
Not open for further replies.
Top