Re: RE24 connection code reworking

[Howard Chu <hyc@symas.com>]

Pierangelo Masarati wrote:
> Howard Chu wrote:
>
> > What kind of system are you running on? Linux / multiprocessor?
>
> No, right now that's just my poor man's laptop (Intel(R) Pentium(R) M
> processor 2.26GHz, UP) running Linux (2.6.18, CentOS 5.2).  Tomorrow I
> can check on better hardware, including SMP (quad Opteron), but I don't
> think that's a problem.  Of course it's using epoll().

Can you try a newer Linux kernel?

The last useful log entries are:

do_syncrep2: rid=001 (-1) Can't contact LDAP server
do_syncrepl: rid=001 retrying (4 retries left)
connection_read(10): no connection!

The do_syncrep2 message is printed at syncrepl.c:1208.
At that point both err and rc are -1. It'll fall into line 1217 and tear down 
the connection, which would remove it from epoll's control. So I don't 
understand why epoll is waking up repeatedly after that point, unless perhaps 
there's a kernel bug in epoll...

--
  -- Howard Chu
  CTO, Symas Corp.           http://www.symas.com
  Director, Highland Sun     http://highlandsun.com/hyc/
  Chief Architect, OpenLDAP  http://www.openldap.org/project/