Bug 43110
Summary: | tasks blocked for more than 120sec with heavy NFS traffic | ||
---|---|---|---|
Product: | File System | Reporter: | warpme (piotr.oniszczuk) |
Component: | NFS | Assignee: | bfields |
Status: | RESOLVED UNREPRODUCIBLE | ||
Severity: | normal | CC: | alan, flo, kernel, trondmy |
Priority: | P1 | ||
Hardware: | All | ||
OS: | Linux | ||
Kernel Version: | 3.0.27 | Subsystem: | |
Regression: | No | Bisected commit-id: |
Description
warpme
2012-04-16 19:36:26 UTC
No idea. Looks kjournald and a bunch of other processes are stuck waiting for IO. What kind of disk is this? Has this only happened this one time, or is it reproduceable? I can reproduce the case: I have set up automounter to mount /data/video via nfsv4 from my video server (otto) (bug is also hit with nfsv3 i think). I try: dd if=/dev/zero of=/data/video/testfile bs=1024k count=100 -> 22MB/s, works great dd if=/dev/zero of=/data/video/testfile bs=1024k count=1000 -> 6MB/s, works great dd if=/dev/zero of=/data/video/testfile bs=1024k count=10000 -> stuck after some time, some processes block. I logged in via ssh and tried to kill dd, so to find the process i did: ps -ef |grep dd -> This command blocks! So i did strace -f ps -ef ... stat("/proc/2528", {st_mode=S_IFDIR|0555, st_size=0, ...}) = 0 open("/proc/2528/stat", O_RDONLY) = 6 read(6, "2528 (plasma-desktop) D 1 2411 2"..., 1023) = 249 close(6) = 0 open("/proc/2528/status", O_RDONLY) = 6 read(6, "Name:\tplasma-desktop\nState:\tD (d"..., 1023) = 910 close(6) = 0 open("/proc/2528/cmdline", O_RDONLY) = 6 read(6, <unfinished ...> Last line comes from CTRL-C i think So something in /proc is blocking? ps -u myusername works btw. and killing dd makes the system usable within some seconds. My guess: If the IO waiting for NFS to write is larger than memory (+swap?), then i hit the bug. System has 4GB memory and 4GB swap, the dd generates 10GB data. More information: Opensuse 12.1, all patches installed, uname -a Linux mamba 3.1.10-1.9-desktop #1 SMP PREEMPT Thu Apr 5 18:48:38 UTC 2012 (4a97ec8) x86_64 x86_64 x86_64 GNU/Linux NFS Server is a Opensuse 11.4, and mount |grep video otto:/video/ on /data/video type nfs4 (rw,relatime,vers=4,rsize=262144,wsize=262144,namlen=255,hard,proto=tcp,port=0,timeo=600,retrans=2,sec=sys,clientaddr=192.168.135.136,minorversion=0,local_lock=none,addr=192.168.135.104) Hi, Unfortunately recently moved to 3.3.7. As soon as i will catch this issue under 3.3.7 - I'll post immediately... -br Is this still happening? I'm not able to reproduce this problem with 3.3.8. But also I have to mention I switch from NFS to CIFS due problems with v.slow speed (1-3MB/s) for MAC->Linux NFS transfers. Strange as Linux->MAC works OK (70-90MB/s) :-( br OK, marking this unreproduceable for now. "problems with v.slow speed (1-3MB/s) for MAC->Linux NFS transfers" That's writes from an OSX client to a Linux server? Possibly worth opening another bug if you want to look into that. "That's writes from an OSX client to a Linux server? Possibly worth opening another bug if you want to look into that." Oh thx for suggestion. I'll do this definitely ! |