[Date Prev][Date Next] [Chronological] [Thread] [Top]

big directories (more than 1M entries)

Hi there,

first i've a question on removing transaction logs..
FAQ-O-Matic and Sleepycat Documentation say, someone should run 
db_archive to determine which Logs can be removed. But now i've added 
about 250k entries into my OpenLDAP 2.1.8 with BDB backend, and i 
noticed that i have to restart (or at least stop) slapd (same for 
slapadd, as i notice) before db_archive shows me unused logs.
Is that intended behaviour?
I really don't like that, 'cause my whole dataset in that specific 
backend has slightly more than 1Million entries, and Transaction log 
would grow to about 4GB of data. That wouldn't be that bad, but I'm 
working on a ramdisk, and on AIX maximum size of ramdisks seems to be 
limited to 2GB.

Next i noticed that slapdindex grows quite large when indexing the whole 
backend, i.e. it grows larger than 256MB, which is the maximum size for 
processes on AIX 32 and 64 Bit systems if the binary has been compiled 
without any -bmaxdata switch. Maybe this should be change to default to 
somethin bigger in the Makefiles, or at least should be mentioned in 
the FAQs? Further information about that can be obtained from [0].

Well, after that worked out (I patched slapindex using dd) I started 
slapd and issued a simple search on an indexed attribute, but i had to 
notice that slapd isn't using the index files. (Not even `fuser` shows 
a process attached to the index files.) After ~2 minutes slapd returns 
the queried entry.
I generated the DB with slapadd and had the index stuff commented out in 
slapd.conf. Could that cause a problem? Running slapindex should have 
created all needed indexes, shouldn't it?

Another topic with respect to problems mentioned above..
I thought of splitting that backend into several smaller parts and 
having them tied back together by slapd.
I vaguely remember someone on the list mentioned a possibility to do 
this without referrals? With a glue backend or something like that? The 
important thing is, that entries have to be queried as
allthought they are internally split up into
Or maybe i can acomplish this with a suffixalias setting? I.e. setting 
the same suffixalias (ou=mail-users,..) for several backends? (Just for 
read, no writes on that alias, of course..)



PS: Migrating from OpenLDAP 1.2.[8|11|12|13] (a little bit of all) to 
OpenLDAP 2.1.8 on AIX 4.3.3 ML 10 on a quad Power4 CPU with 4 GB of 
RAM. (p660) Compiled with xlc_r, threaded and BerkeleyDB 4.1.24 Libs.
Unix? What's that? Is that like Linux?