slow access, strange errors

[Geoff Silver <geoff@uslinux.net>]

I'm running OpenLDAP 2.0.23 with libnss-ldap 186-1 under Debian Linux.  I
have several LDAP servers/replicas.  Strangely, an 'ls -l /home' on the
system takes about 6.5 seconds to run on ~90 entries when querying a
remote server with a 2.4.18 kernel, but 11.5 to do the same on the local
slapd server with a 2.2.18 kernel.  Either way that seems to take a very
long time, but it's particular disturbing that two identical systems, both
Athlon XP 1700+ with plenty of RAM perform so much differently.

I've tried some slapd.conf tuning, including setting:
	cachesize       100000
	dbcachesize     20000000
	index default pres,eq
	index cn,sn,mail,givenName pres,eq,sub

but no luck.  What is throwing me off in particular is a log message which
says: slapd[23804]: deferring operation.  If I run in -d4095 (super high
debugging mode), the only problem I'm seeing is:

ber_get_next
ldap_read: want=1 error=Resource temporarily unavailable
ber_get_next on fd 9 failed errno=11 (Resource temporarily unavailable)
do_search
ber_scanf fmt ({aiiiib) ber:
ber_dump: buf=0x080d6088 ptr=0x080d608b end=0x080d614c len=193

Has anyone else experienced anything like this?  Is there any reason why
slapd for pam/nss use shouldn't be nearly as fast as local /etc files,
particularly when the LDAP server is on the local system?  Thanks.


-- 
Geoff Silver					<geoff at uslinux dot net>
"If Bill Gates had a nickel for every time Windows crashed...
	Oh wait, he does"