Page 2 of 8 FirstFirst 1234 ... LastLast
Results 11 to 20 of 72

Thread: large NFS copy locks up/hang client with large files (again) (lucid)

  1. #11
    Join Date
    Jan 2007
    Beans
    33

    Re: large NFS copy locks up/hang client with large files (again) (lucid)

    Hello,

    Same here:
    Client is Lucid 64bits
    Server is freenas

    I would say it was working fine when I was in Ubuntu 9.10 but as I've updated my freenas server recently, I am not 100% sure it is related to Ubuntu or freenas.

    Any, it sux as samba share or ftp are working fine (& faster!) both of them while nfs is just freezing my computer (client)

  2. #12
    Join Date
    Apr 2006
    Beans
    136

    Re: large NFS copy locks up/hang client with large files (again) (lucid)

    I've had this problem on numerous Ubuntu distros for over a year now... very frustrating. Will be trying to see if fedora suffers from similar problems tonight.

    Has this specific bug been filed? I'm also on gigabit.

  3. #13
    Join Date
    Jan 2007
    Beans
    33

    Re: large NFS copy locks up/hang client with large files (again) (lucid)

    Up the thread to see if anyone has an idea ?

  4. #14
    Join Date
    Dec 2006
    Location
    Tommy Douglas' Mouseland
    Beans
    74

    Re: large NFS copy locks up/hang client with large files (again) (lucid)

    Is there any graphical utility that can record overall system throughput; uplink + downlink; that people can recommend?

    Ideal if it could save/export the data too.

    I've used iperf, system utility/monitor, iptraf, etc. for looking at near real-time throughput, but it'd be nice to be able to see actually NFS throughput recordable in graph form...

    I'd use this for benchmarking + troubleshooting purposes on this issue.

    Thanks

  5. #15
    Join Date
    Apr 2006
    Beans
    136

    Re: large NFS copy locks up/hang client with large files (again) (lucid)

    I've noticed that I only have problems when writing to my nfs mounts, and not reading from them.

    excerpt from fstab mount:

    Code:
    nfs		  tcp,intr	  0	0
    EDIT: I have complete lockups from wireless to NFS writes... I don't think its a gigabit issue.
    Last edited by zackiv31; June 13th, 2010 at 07:25 PM.

  6. #16
    Join Date
    Jan 2007
    Beans
    33

    Re: large NFS copy locks up/hang client with large files (again) (lucid)

    The simple fact to extract a big archive (rar/zip) from my local HDD to my local HDD make my computer freezing while the unrar is processing (it take about 25% of my CPU ressources) and I last in about 1 minute for a couple of GB archive which seems to be fine.

    I do not have anything strange in dmesg.

    Any help to find what's going on ?

  7. #17
    Join Date
    Dec 2006
    Location
    Tommy Douglas' Mouseland
    Beans
    74

    Re: large NFS copy locks up/hang client with large files (again) (lucid)

    Some more info; no resolution.

    Systems
    -----------------------
    Server = Atom 330 mobo [ECS 945GCD-M (V1.0)], 4GB DDR2, Mythbuntu 9.10, NFSv4

    Client = Core 2 e8400, ubuntu 9.04, NFSv4, 4 GB DDR2

    Network = ALL Gigabit, 2 x 8 port DLink DGS-1008D

    Problem
    -----------------------
    Using NFS v4 standard mount options and combo of other tests an NFS write of a large file; 8.4 GB in this test case; will start off with a very large burst of 60+ MB/s and then more or less level off at an anemic 12 MB/s.

    Looking at the throughput graphs, however, this is a very far from stable 12 MB/s; throughput goes up and down like a bride's pajamas...OVERALL NFS WRITE SPEED SUCKS.

    See attached pictures; bursty_2.png + bursty_3.png

    What I Have Done To Try And Isolate The Problem
    -----------------------
    server = Load is very high; 15+ when doing the NFS xfer but CPU, IOWAIT, MEM are NOT conspicuously high as well; so I checked things like nfsstat -rc, iostat, netstat -ic, iotop, etc and don't see obvious issues.

    See attached text file; bursty_txt.txt for details.

    What I Have Done To Try Fix The Problem
    -----------------------
    * note: have been using default mount of:
    sudo mount -t nfs4 192.168.YY.XX:/temp0 /mnt/nt_tmp/
    unless otherwise noted.

    -- increased amount of nfsd threads on server from 8 to 32 = no change
    -- mtu on both ends from 1500 to 9000 = no change
    -- mount option rsize,wsize from default of 524288 to 262144 = no change
    -- mount option rsize,wsize from default of 524288 to 1048576 = maxes out at 524288, no change

    Overall Results
    -----------------------
    -- Xfering a 8.4 GB file over Gigabit NFSv4 = 14m53s

    -- Compared to Samba it is SLOWER; Samba consistent 16 MB/s and 9m30s for same file

    -- What I'll try next is a) 100 Mbits switches on each end; b) faster server...eventually.

    Closing
    -----------------------
    -- At this point I am wondering if the atom330 server is just too slow?! Units seems to work well; browser, etc still very responsive even when the load says 15+ for the NFS xfers...

    -- Any one have any other ideas for me to test?

    Thanks
    Attached Images Attached Images
    Attached Files Attached Files

  8. #18
    Join Date
    Dec 2006
    Location
    Tommy Douglas' Mouseland
    Beans
    74

    Re: large NFS copy locks up/hang client with large files (again) (lucid)

    FYI. From what I can tell the cpu + mdadm combo is NOT the issue / bottleneck...

    <user>@atom330:~$ time dd if=/dev/zero of=/home/<user>/raid5/test.dd bs=8k count=1000000
    1000000+0 records in
    1000000+0 records out
    8192000000 bytes (8.2 GB) copied, 137.285 s, 59.7 MB/s

    real 2m17.716s
    user 0m1.420s
    sys 1m21.340s

    <user>g@atom330:~$ time dd if=/dev/zero of=/home/<user>/raid5/test.dd bs=8k count=1000000
    1000000+0 records in
    1000000+0 records out
    8192000000 bytes (8.2 GB) copied, 148.631 s, 55.1 MB/s

    real 2m31.178s
    user 0m1.590s
    sys 1m33.970s

    <user>g@atom330:~$ time dd if=/dev/zero of=/home/<user>/raid5/test.dd bs=8k count=1000000
    1000000+0 records in
    1000000+0 records out
    8192000000 bytes (8.2 GB) copied, 143.58 s, 57.1 MB/s

    real 2m25.909s
    user 0m1.620s
    sys 1m32.640s

    This same test with an 8.4 GB file over NFSv4 took 14m53s.

    dd testing the load never went above 2.5 either...was 15+ for the NFS transfers; MUST be something on the networking side?

    Funny thing is an iperf test to this server averages 840 Mbits/sec -OR- 100+MB/s using TCP so I'm not sure why these file xfers are so terrible.

  9. #19
    Join Date
    Jan 2007
    Beans
    33

    Re: large NFS copy locks up/hang client with large files (again) (lucid)

    on my side, it seems that my system has problems when reading large local files as :
    1- I am able to read from my NFS share with good speed (30-40MB/s) and write it to my local HDD
    2- When I try to copy a file from my local HDD, My computer freeze randomly
    3- Same Computer, Same OS, Same files (copied/read to a NTFS partition) I am experiencing the exactly same behavior
    4- Same Computer, Same OS, Same files but using CIFS produce the same behavior
    4- Running on the same Computer, but from Windows XP, I do not get any freeze.

    I'm am now trying with an Ubuntu LiveCD 9.10 and i'll keep you updated...

  10. #20
    Join Date
    Jan 2007
    Beans
    33

    Re: large NFS copy locks up/hang client with large files (again) (lucid)

    No performances issues with my LiveCD 9.10

    It seems to be definitly a problem with 10.04 on my computer...
    Anyone has an idea ?

Page 2 of 8 FirstFirst 1234 ... LastLast

Tags for this Thread

Bookmarks

Posting Permissions

  • You may not post new threads
  • You may not post replies
  • You may not post attachments
  • You may not edit your posts
  •