Re: nfs problems

["Christoph Egger" <Christoph_Egger%gmx.de@localhost>]

> 
> > On Tue, Dec 08, 2009 at 06:07:32PM +0100, Christoph Egger wrote:
> >  > I observe strange hangs with nfs on amd64.
> >  > 
> >  > 'ls' works fine but 'sync' blocks the shell and
> >  > isn't interruptable with CTRL-c.
> >  > 
> >  > I can provide more information when you tell me
> >  > which ones you want to have.
> >  > I don't know where to look at.
> > 
> > A good start for this or any other hang is to run ps -alx and
> > examine the WCHAN column.
> > 
> > If, as is likely, everything is waiting on 'tstile', find the
> > related process that isn't waiting on tstile that all the
> > others are queueing up behind. Usually there is one.
> 
> UID  PID PPID  CPU PRI NI   VSZ   RSS WCHAN   STAT TTY         TIME
> COMMAND
>   0    0    0    0 125  0     0 25520 uvm     DKl  ?        0:57.28
> [system]
>   0    1    0  240  85  0 10640  1084 wait    Is   ?        0:00.02 init 
>   0   72    1   79  43  0 13704  1152 parked  Isl  ?        0:00.02
> xenbackendd
>   0  139    1    0  85  0 21164  1600 kqueue  Ss   ?        0:00.23
> /usr/sbin/s
>   0  306    1  306  85  0 43160  1992 select  Is   ?        0:00.01
> /usr/sbin/s
>   0  356    1    0  85  0  8572  1056 nanoslp Ss   ?        0:00.25
> /usr/sbin/c
>   0  383  306    0  85  0 62292  4060 select  Ss   ?        0:05.25 sshd:
> root@
>   0  460    1    0  85  0 19168  2256 kqueue  Ss   ?        0:00.56
> /usr/libexe
>  12  472  460    0  85  0 19276  2392 kqueue  I    ?        0:00.31 qmgr
> -l -t 
>   0  474    1  671  85  0 12752  1048 kqueue  Is   ?        0:00.00
> /usr/sbin/i
>   0  554  571 4097  85  0 91136 17172 select  Il   ?        0:45.20
> /usr/pkg/bi
>   0  563    1    0  85  0 11776  1632 select  I    ?        0:01.65
> xenstored -
>   0  564    1    0  85  0 19012  1164 select  Il   ?        0:00.02
> xenconsoled
>   0  571    1  805  85  0 68608  4624 wait    I    ?        0:00.02
> /usr/pkg/bi
>   0  815  554  413  85  0 52860  8524 genput  Dl   ?       29:34.16
> /usr/local/
>  12 2053  460    0  85  0 19212  2308 kqueue  S    ?        0:00.04 pickup
> -l -
>   0  134  438    0  85  0 10788  1164 wait    I+   ttyp0    0:00.01
> /bin/sh ./n
>   0  438  383    0  85  0  6600  1244 pause   Is   ttyp0    0:00.03 -ksh 
>   0  613  622    0   0  0     0     0 -       Z+   ttyp0    0:00.00
> (python2.5)
>   0  622  134  303  85  0 16940  1228 select  Sl+  ttyp0    0:05.52
> /usr/local/
>   0  387  497    0  85  0  6604  1288 pause   S    xencons  0:01.55 -ksh 
>   0  497    1  483  85  0 52740  2952 wait    Is   xencons  0:00.05 login 
>   0 1410  387    0  42  0 10672  1000 -       O+   xencons  0:00.01 ps
> -alx 
> 
> 
> There is no 'tstile'. But there is one 'parked'. What
> does that mean ?

In that situation when I do 'sync' it is in 'tstile':


UID  PID PPID  CPU PRI NI   VSZ   RSS WCHAN   STAT TTY         TIME COMMAND
  0    0    0    0 125  0     0 25596 uvm     DKl  ?        0:57.58 [system]
  0    1    0    0  85  0 10640  1084 wait    Ss   ?        0:00.02 init 
  0   72    1   79  43  0 13704  1152 parked  Isl  ?        0:00.02 xenbackendd
  0  139    1    0  85  0 21164  1600 kqueue  Ss   ?        0:00.23 /usr/sbin/s
  0  306    1    0  85  0 43160  1992 select  Ss   ?        0:00.01 /usr/sbin/s
  0  356    1    0  85  0  8572  1056 nanoslp Is   ?        0:00.25 /usr/sbin/c
  0  460    1    0  85  0 19168  2256 kqueue  Is   ?        0:00.57 /usr/libexe
 12  472  460    0  85  0 19276  2392 kqueue  I    ?        0:00.31 qmgr -l -t 
  0  474    1  671  85  0 12752  1048 kqueue  Is   ?        0:00.00 /usr/sbin/i
  0  554  571 4097  85  0 91136 17172 select  Il   ?        0:45.71 /usr/pkg/bi
  0  563    1    0  85  0 11776  1632 select  I    ?        0:01.65 xenstored -
  0  564    1    0  85  0 19012  1164 select  Il   ?        0:00.02 xenconsoled
  0  571    1  805  85  0 68608  4624 wait    I    ?        0:00.02 /usr/pkg/bi
  0  815  554  413 117  0 52860  8524 genput  Dl   ?       29:34.16 /usr/local/
  0 1955  306    0  85  0 62292  4128 select  Ss   ?        0:00.05 sshd: root@
 12 2053  460    0  85  0 19212  2308 kqueue  I    ?        0:00.05 pickup -l -
  0  613  622    0   0  0     0     0 -       Z    ttyp0 -  0:00.00 (python2.5)
  0  622    1    0   0  0     0     0 -       DE   ttyp0 -  0:00.00 (xenconsole
  0 1724 1823    0 117  0  5368   604 tstile  D+   ttyp0    0:00.01 sync 
  0 1823 1955    0  85  0  6600  1252 pause   Ss   ttyp0    0:00.03 -ksh 
  0  387  497    0  85  0  6604  1288 pause   S    xencons  0:01.57 -ksh 
  0  497    1  483  85  0 52740  2952 wait    Is   xencons  0:00.05 login 
  0 2729  387 1024  42  0 10672  1000 -       O+   xencons  0:00.01 ps -alx 

So something bad already happens before a process ends up
in tstile hang.

Christoph