Re: r247839: broken pipe - for top, sudo and ports

On 03/06/13 13:29, Dimitry Andric wrote:
> On 2013-03-05 20:59, Hartmann, O. wrote:
> ...
>> A "truss top" reveals this, is this of help?
> ...
>> socket(PF_LOCAL,SOCK_STREAM,0)                   = 4 (0x4)
>> connect(4,{ AF_UNIX "/var/run/nscd" },15)        = 0 (0x0)
>> fcntl(4,F_SETFL,O_NONBLOCK)                      = 0 (0x0)
> ...
>> sendmsg(0x4,0x7fffffffd290,0x0,0x1,0x1,0x0)      ERR#32 'Broken pipe'
> 
> Maybe your nscd might have died?  It is successfully opening and
> connecting to nscd's unix socket, but gets an EPIPE when it tries to
> send data.


I tried "service nscd stop", stopping nscd, but I guess I also have to
change configuration /etc/nsswitch.conf then to make that effective.

Even with stopped nscd, I have those weird and strange "broken pipes" as
I reported (a little bit unsensible, I'm sorry).

I use TEMPFS for /var/run, /tmp. I also use nscd, OpenLDAP. This works
well with r247545, but with ~r247839 the strange behaviour occurs.

My gateway system dies due loosing connection sporadically because of
dying services.

Could this "socket problem" also affect any kind of user verification,
say, when a process starts up, but can not find its ID, like root, wheel
or ldap (for OpenLDAP) or nagios?

I reverted my sources to r247479 now and try to stick with those until
the problem has been identified by more bright people and developers.
I'm only an "advanced user", if ever, my abilities to debug are very
limited and for that I'm sorry. I can only provide "detailed" reports of
my desperate search for a solution.

Thanks and regards,

Oliver