crazy network problem -- sleep?

[Adrian Penisoara]

Hi,

2009/3/1 stutiredboy <stutiredboy at gmail.com>

> Adrian Penisoara 写道:
> > Hi,
> >
> > 2009/3/1 stutiredboy <stutiredboy at gmail.com
> > <mailto:stutiredboy at gmail.com>>
> >
> >     hi,all:
> >
> >     recently , we found a problem that , some server which with FreeBSD
> >     6.2 may stop networking after
> >     days of services, it is so difficulty express this status, e, such as
> >     sleep? i can not ping this server, tracert is also
> >     failure, and other services ,network/server seems to be down... but
> >     while i insert the keyboard/mouse/crt in ,
> >     everything is ok, such as a ring to call the system to wake-up... no
> >     errors found in dmesg or /var/log/messages,
> >     such as nothing happened...orz，our unlimit args as follow:
> >
> >     cputime unlimited
> >     filesize unlimited
> >     datasize 2088152 kbytes
> >     stacksize 65536 kbytes
> >     coredumpsize unlimited
> >     memoryuse unlimited
> >     vmemoryuse unlimited
> >     descriptors 655000
> >     memorylocked unlimited
> >     maxproc 5547
> >     sbsize unlimited
> >
> >     plz give me some suggestion to trace the problem or solve it .
> >     thanks a
> >     lot !
> >
> >
> > First you need to check whether packets are reaching the system and
> > whether the response packets are leaving it with tcpdump(8) on the
> > appropriate interface.
> >
> > If you see the packets coming but not coming out then you need to
> > check the status of your TCP/IP stack -- e.g. post the output of
> > "netstat -m" and "netstat -s". Also check dmesg output for any errors.
> >
> > If you are using firewall(s) (ipfw/ipfilter/pf) then check the rules
> > and the firewall states status; try temporarily disabling the firewall
> > (ipfw disable firewall / ipf -D / pfctl -d) and check whether your
> > system becomes reachable.
> >
> > Check CPU load, interrupts, packet stats (systat -vmstat / systat
> > -ifstat).
> >
> > Regards,
> > Adrian Penisoara
> > EnterpriseBSD
> hi, thank you for your help, i can only ssh to the system, so while the
> network is down, i can not login in
>
> yes, we have ipfw on , these rules have used a long time, and many other
> severs do not have this problem,
>
> this sistutation happened scarce, in different machine rooms, different
> isps, different servers, such as hp\dell.
>
> while the network usable, the cpu and system load is normal, i have
> checked dmesg and syslog, nothing
>
> special found , where our colleague in machine room insert the keyboard
> or mouse in ,then the network
>
> recover , it is so difficult to catch the ouput of tcpdump or other
> messages form netstat , the moment you
>
> have the keyboard connected , the network recover... thanks very much !


You mean that once you plugin a keyboard the machine recovers ? That's
weird. What if you leave the keyboard connected all the time ?

Try to attach one of the machines which reproduces the problem to a KVM
switch (Keyboard/Video/Mouse switcher). If you are able to reproduce again
the problem then you might have more chances to catch it.

Regards,
Adrian.