SLAPD accepts connection, then hangs

[Dave Horsfall <daveh@ci.com.au>]

OpenLDAP 2.2.26, FreeBSD 4.10-STABLE.

Yeah, I know it's old, but I'm hoping for an answer along the lines of 
"Yeah, we fixed that in 2.2.X".  I can't find anything like it in CHANGES 
in 2.2.30, nor in the ITS.  Going to 2.3.X is not yet an option (unless of 
course that's where the fix is).

The symptom is that SLAPD accepts a connection, then nothing further 
happens.  This happens around once a month, making it difficult to nail.  
Because the connection has been accepted, the client does not bother 
trying the backup server.

The log looks like this:

Jan 19 15:34:46 guido slapd[7718]: conn=177196 fd=438 ACCEPT from IP=XXX:3331 (IP=0.0.0.0:389)
Jan 19 15:34:55 guido slapd[7718]: conn=177197 fd=439 ACCEPT from IP=XXX:1326 (IP=0.0.0.0:389)

and nothing else.

Truss shows:

gettimeofday(0x111899d4,0x0)                     = 0 (0x0)
poll(0x10d75800,0x1,0x0)                         = 0 (0x0)

over and over.

GDB backtrace:

#0  0x283d8d44 in __sys_poll () from /usr/lib/libc_r.so.4
#1  0x283d8290 in _thread_kern_sched_state_unlock () from /usr/lib/libc_r.so.4
#2  0x283d7c55 in _thread_kern_scheduler () from /usr/lib/libc_r.so.4
#3  0x0 in ?? ()

I'll rebuild with a better symbol table next time :-(

Ring any bells with anyone?

-- 
Dave Horsfall  DTM  VK2KFU  daveh@ci.com.au  Ph: +61 2 9552-5509 (d) -5500 (sw)
Corinthian Engineering P/L,  54/19-21 Pirrama Rd, Pyrmont, NSW 2009,  Australia