named problem?

Fedora Core 5
kernel-smp-2.6.18-1.2239.fc5
bind-utils-9.3.3-0.2.rc2.fc5
bind-libs-9.3.3-0.2.rc2.fc5
bind-chroot-9.3.3-0.2.rc2.fc5

At some chaotic interval my box will hang to the extent
that I cannot even 'ls -l /tmp' (may not pertain to anything).
I cannot 'shutdown -fr now' the box until I kill named. Once
I restart the box things are ok until the next time this happens.
I am running 'top' and do not see named spinning. I am going to
cause named to do more logging (named -d 3?) to see what's going
on. Any suggestions to what might be causing the problems?

Mike

> At some chaotic interval my box will hang to the extent
> that I cannot even 'ls -l /tmp' (may not pertain to anything).
> I cannot 'shutdown -fr now' the box until I kill named. Once

What sort of networking might be involved in "ls -l"? I can think of a
couple:
- remote mounted filesystem in your search path
- using NIS for passwd or group

On 2007-01-14, Allen McIntosh <(E-Mail Removed)> wrote:
>> At some chaotic interval my box will hang to the extent
>> that I cannot even 'ls -l /tmp' (may not pertain to anything).
>> I cannot 'shutdown -fr now' the box until I kill named. Once
>
> What sort of networking might be involved in "ls -l"? I can think of a
> couple:
> - remote mounted filesystem in your search path
> - using NIS for passwd or group

None that I can think of, the ls above is of /tmp. Yet, when I kill
named the ls works again. No NIS, though I am moving to ldap (OpenLDAP).

Mike

Mike wrote:
> On 2007-01-14, Allen McIntosh <(E-Mail Removed)> wrote:
>>> At some chaotic interval my box will hang to the extent
>>> that I cannot even 'ls -l /tmp' (may not pertain to anything).
>>> I cannot 'shutdown -fr now' the box until I kill named. Once
>> What sort of networking might be involved in "ls -l"? I can think of a
>> couple:
>> - remote mounted filesystem in your search path
>> - using NIS for passwd or group
>
> None that I can think of, the ls above is of /tmp. Yet, when I kill
> named the ls works again. No NIS, though I am moving to ldap (OpenLDAP).

Anything weird in /etc/nsswitch.conf?
I suppose you could use strace and see where the ls hangs. On my system
(vanilla FC3) I see ls doing this:

socket(PF_FILE, SOCK_STREAM, 0) = 3
fcntl64(3, F_GETFL) = 0x2 (flags O_RDWR)
fcntl64(3, F_SETFL, O_RDWR|O_NONBLOCK) = 0
connect(3, {sa_family=AF_FILE, path="/var/run/nscd/socket"}, 110) = -1
ENOENT (No such file or directory)
close(3) = 0


which looks like it might be related to your problem. I don't run
named, so the connect fails. I have no idea why it is doing this, but I
have a suspicion that YMMV on this one.

On 2007-01-14, Allen McIntosh <(E-Mail Removed)> wrote:
> Mike wrote:
>> On 2007-01-14, Allen McIntosh <(E-Mail Removed)> wrote:
>>>> At some chaotic interval my box will hang to the extent
>>>> that I cannot even 'ls -l /tmp' (may not pertain to anything).
>>>> I cannot 'shutdown -fr now' the box until I kill named. Once
>>> What sort of networking might be involved in "ls -l"? I can think of a
>>> couple:
>>> - remote mounted filesystem in your search path
>>> - using NIS for passwd or group
>>
>> None that I can think of, the ls above is of /tmp. Yet, when I kill
>> named the ls works again. No NIS, though I am moving to ldap (OpenLDAP).
>
> Anything weird in /etc/nsswitch.conf?
> I suppose you could use strace and see where the ls hangs. On my system
> (vanilla FC3) I see ls doing this:
>
> socket(PF_FILE, SOCK_STREAM, 0) = 3
> fcntl64(3, F_GETFL) = 0x2 (flags O_RDWR)
> fcntl64(3, F_SETFL, O_RDWR|O_NONBLOCK) = 0
> connect(3, {sa_family=AF_FILE, path="/var/run/nscd/socket"}, 110) = -1
> ENOENT (No such file or directory)
> close(3) = 0
>
>
> which looks like it might be related to your problem. I don't run
> named, so the connect fails. I have no idea why it is doing this, but I
> have a suspicion that YMMV on this one.

Doh, running 'strace ls -l /tmp' to see where ls is failing is an excellent
idea. I'll try that next time ls fails. I have the debugging increased on
named for now.

Mike

In news:5kCqh.72$(E-Mail Removed),
Mike <(E-Mail Removed)> wrote:

>>>>> At some chaotic interval my box will hang to the extent
>>>>> that I cannot even 'ls -l /tmp' (may not pertain to anything).
>>>>> I cannot 'shutdown -fr now' the box until I kill named.
....
> Doh, running 'strace ls -l /tmp' to see where ls is failing is an
> excellent idea. I'll try that next time ls fails.

What'w wrong with this picture?

On 2007-01-15, patrick <(E-Mail Removed)> wrote:
> In news:5kCqh.72$(E-Mail Removed),
> Mike <(E-Mail Removed)> wrote:
>
>>>>>> At some chaotic interval my box will hang to the extent
>>>>>> that I cannot even 'ls -l /tmp' (may not pertain to anything).
>>>>>> I cannot 'shutdown -fr now' the box until I kill named.
> ...
>> Doh, running 'strace ls -l /tmp' to see where ls is failing is an
>> excellent idea. I'll try that next time ls fails.
>
> What'w wrong with this picture?

If I knew maybe I could fix the problem sooner. I know when the
event happens that 'ls -l /tmp' freezes and I can execute the same
command after I kill named.

Mike