[Samba] Samba 3.3.12/3.4.7/3.5.2 on "ancient" server stops responding/dies

[Luca Olivetti]

Al 02/05/10 10:43, En/na Luca Olivetti ha escrit:

>> After a long round of bisecting I found that the last working version
>> (apart from the 1 minute delay) is 3.2.5.
>> I didn't find anything like this about 3.2.6 in the mailing list :-/
>
> If I use "smbclient -L localhost" instead of "smbclient -L hp", one of
> the smbd child dies silently (at least the master process complains that
> it doesn't find its pid any longer) right after searching for the domain
> record in ldap (with apparently a successful result).
> Thinking that maybe the problem was related with ldap client libraries,
> I updated ldap to 2.4.19 (only the client libraries, didn't touch the
> server) but the problem persisted, so maybe that's just a red herring.
>
> Any hint on how to debug this further?

I attached strace to the samba processes and this is what I see:

1) in the level 10 log:

[2010/05/02 17:33:38,  2, pid=2379] lib/messages_local.c:message_notify(270)
   message to process 2414 failed - No such process

the last message from pid 2414 is:
[2010/05/02 17:33:38,  5, pid=2414] lib/smbldap.c:smbldap_search_ext(1200)
   smbldap_search_ext: base => 
[sambaDomainName=WETRON,sambaDomainName=WETRON,dc=wetron,dc=es], filter 
=> 
[(&(objectClass=sambaTrustedDomainPassword)(sambaDomainName=WETRON))], 
scope => [2]


2) and with strace I see that pid 2414 got a SIGPIPE writing to fd 8

[pid  2414] write(23, "  smbldap_search_ext: base => [s"..., 189) = 189
[pid  2414] rt_sigaction(SIGALRM, {0x803fa895, [ALRM], SA_RESTORER, 
0x40328718}, {SIG_IGN}, 8) = 0
[pid  2414] alarm(15)                   = 0
[pid  2414] time(NULL)                  = 1272814418
[pid  2414] rt_sigaction(SIGALRM, {0x803fa895, [ALRM], SA_RESTORER, 
0x40328718}, {0x803fa895, [ALRM], SA_RESTORER, 0x40328718}, 8) = 0
[pid  2414] alarm(15)                   = 15
[pid  2414] write(8, "0\204\0\0\0\5\2\1\23B\0", 11) = -1 EPIPE (Broken pipe)
[pid  2414] --- SIGPIPE (Broken pipe) @ 0 (0) ---
Process 2414 detached

I don't see where pid 2414 opens fd 8, either strace didn't catch it or 
it was inherited from the parent, but I suppose it's the connection with 
the ldap server.

Bye
-- 
Luca Olivetti
Wetron Automatización S.A. http://www.wetron.es/
Tel. +34 93 5883004 (Ext.133)  Fax +34 93 5883007