Re: NFS mount very very slow
Dear John et al,
(Firstly, sorry about the previous double posting. At first I posted
using a different e-mail address to that under which I had been
subscribed, so I resent it using the correct address.)
On Sun, 27 Jun 2004 01:53 pm, John Summerfield wrote:
> James Sinnamon wrote:
> >Dear Paulo et al,
> >
> >On Sun, 27 Jun 2004 07:12 am, Paulo Silva wrote:
> >>A Sex, 2004-06-25 às 12:25, James Sinnamon escreveu:
> >>>Dear list subscribers,
> >>>
> >>>At first I thought my command:
> >>>
> >>> mount -t nfs 192.168.0.6:/etc /mnt/nfs
> >>>
> >>>... had failed. It seemed to have hanged. I tried to kill it with
> >>>Ctrl^C, then with 'kill -SIGKILL <procid>', killing the Konsole tabbed
> >>>terminal ... but with no luck.
> >>>
> >>>The process just would not die.
> >>>
> >>>Eventually, I found, to my surprise, that that the 'mount' command
> >>>had not only withstood all my attempts to smother it, but it also
> >>>succeeded after all. I don't know whether it took 15 minutes
> >>>or two hours, but whatever the time lag was, it it had taken far too
> >>>long. can anyone tell me how to work out what the problem could be?
> >>>
> >>>The entry in 192.168.0.6:/etc/exports is:
> >>>
> >>> /etc 192.168.0.2(ro)
> >>>
> >>>.... where 192.168.0.2 is the nfs client.
> >>
> >>Do you have portmap running?
> >
> >Yes, thank you:
> >
> > greenhouse:~# ps auxwww | grep portmap
> > daemon 512 0.0 0.0 1604 544 ? Ss Jun26 0:00 /sbin/portmap
> >
> >I am running rpc.mountd and rpc.nfsd with the --debug call flag set:
> >
> > greenhouse:~# ps uxwww | grep rpc.nfsd
> > root 2428 0.0 0.0 1940 804 ? Ss 12:29 0:00 /usr/sbin/rpc.nfsd \
> > --debug call
> >
> > greenhouse:~# ps uxwww | grep rpc.mountd
> > root 2421 0.0 0.0 1936 816 ? S 12:27 0:00 /usr/sbin/rpc.mountd \
> > --debug call
> >
> >This causes debug output, the meaning of which is not
> >completely clear to me:
> >
> > greenhouse:/var/log# tail -25f debug
> > Jun 27 12:30:45 greenhouse mountd[2421]: mnt [1 104/6/27 \
> > 12:29:45 moominvalley 0.0+0,1,2,3,4,6,10]
> > Jun 27 12:30:45 greenhouse mountd[2421]: ^I/
> > Jun 27 12:30:45 greenhouse mountd[2421]: ^Imount res = 0
> > Jun 27 12:30:45 greenhouse nfsd[2428]: getattr [1 70/1/1 \
> > 12:47:12 moominvalley 0.0+0,1,2,3,4,6,10]
> > Jun 27 12:30:45 greenhouse nfsd[2428]: ^I58800002 00
> > Jun 27 12:30:45 greenhouse nfsd[2428]: result: 0
> > Jun 27 12:30:45 greenhouse nfsd[2428]: statfs [1 70/1/1 \
> > 12:47:12 moominvalley 0.0+0,1,2,3,4,6,10]
> > Jun 27 12:30:45 greenhouse nfsd[2428]: ^I/
> > Jun 27 12:30:45 greenhouse nfsd[2428]: result: 0
> > Jun 27 12:46:30 greenhouse nfsd[2428]: statfs [1 70/1/1 \
> > 13:02:57 moominvalley 1525.503+503,505,510]
> > Jun 27 12:46:30 greenhouse nfsd[2428]: ^I/
> > Jun 27 12:46:30 greenhouse nfsd[2428]: result: 0
> > Jun 27 12:46:30 greenhouse nfsd[2428]: getattr [1 \
> > 70/1/1 13:02:57 moominvalley 1525.503+503,505,510]
> > Jun 27 12:46:30 greenhouse nfsd[2428]: ^I/
> > Jun 27 12:46:30 greenhouse nfsd[2428]: result: 0
> >
> >
> >It does show evidence of a 16 minute delay between the issuing of the
> >command:
> >
> > mount -t nfs 192.168.0.6:/ /mnt/nfs
> >
> >... on the client host, 192.168.0.2(aka moominvalley), and the root
> >file system of the server host (192.168.0.6 aka greenhouse) finally
> >being mounted.
> >
> >The same happens regardless of what directoryI choose on the
> >server machine.
> >
> >Sun NFS lock manager(?) in kernel?
> >-----------------------------------------------------
> >
> >My bootup script (/etc/init.d/nfs-common) causes /sbin/rpc.lockd to run,
> >but /sbin/rpc.lockd seems to detect there already is an equivalent to
> >lockd (Sun NFS lock manager NLM(?)) in the kernel, and so it ceases
> >execution.
> >
> >I have set /proc/sun/sunrpc/nlm_debug to "1" with no apparent effect?
> >So how can I verify that there is NLM in my kernel, and that it is
> >working?
> >
> > ... and if NLM/lockd is working, what else could be slowing down
> >the NFS mounting of the server file systems?.
>
> I didn't think to look here b4:
> ns:~# lsmod
> ...
> nfsd 179776 1
> exportfs 6496 1 nfsd
> lockd 63176 2 nfsd
> sunrpc 138600 2 nfsd,lockd
> ....
>
> Again, in case you forgot:
> ns:~# uname -r
> 2.6.3-1-686
> ns:~#
> My startup script is nfs-kernel-server.
Thanks. I installed the kernel-server package, but it seems to still
be just as slow.
greenhouse:/var/log# lsmod | grep nfsd
nfsd 61200 8 (autoclean)
lockd 42544 1 (autoclean) [nfsd]
sunrpc 58688 1 (autoclean) [nfsd lockd]
(curiously, I don't see 'exportfs' in my modules, although
/usr/sbin/exportfs exists. Is that an issue?)
Maybe nfs-kernel-server has to be configured diffently? (I haven't
thoroughly read up on it yet.)
I am using a 2.4 kernel: 2.4.25-1-386.
I am quite happy to upgrade if I were to feel confident that it would
work correctly on my testing/sarge system, however I would have
thought that nfs, whether run as a user process or somehow as part
of the kernel, should still work whether it is with a 2.4 kernel or a
2.6 kernel.
So essentially, I have not moved much further towards being
able to find the cause to the problem or the solution.
Any possible further suggestions?
thanks again.
regards,
James
--
James Sinnamon
jps at westnet com auStralia
ph +61 412 319669, +61 2 95692123, +61 2 95726357
Reply to: