Re: [389-users] Tuning 389 DS

Juan Asensio Sánchez wrote:

2010/8/3 Rich Megginson <rmeggins@redhat.com mailto:rmeggins@redhat.com>

Juan Asensio Sánchez wrote:
>
> 2010/8/2 Rich Megginson <rmeggins@redhat.com
<mailto:rmeggins@redhat.com> <mailto:rmeggins@redhat.com
<mailto:rmeggins@redhat.com>>>
>
>     Juan Asensio Sánchez wrote:
>     > Hi
>     >
>     > I am trying to tune the performance of the Directory
Server. We have
>     > increased the memory for the database cache and for each
database
>     > entry cache. These are the new values:
>     >
>     > cn=config, cn=ldbm database, cn=plugins, cn=config
>     > nsslapd-dbcachesize: 838860800 (~800MB)
>     >
>     > cn=*,cn=ldbm database, cn=plugins, cn=config
>     > nsslapd-cachememsize: 125829120 (~120MB)
>     >
>     > We have 27 databases, and the servers have 16 GB of RAM,
so the
>     server
>     > should be able to handle all that memory (800 + 120*27 =
>     4040MB). But
>     > when I go to the monitoring section of the management
console, the
>     > database cache says the hit ratio is 99% (this is OK according
>     to the
>     > documentation, near 100%), but the entry cache is 0%, that is
>     very far
>     > for 100% that the documentation recomends (see screenshots
>     attached).
>     > Am I confused or the configuration is not correct?
>     When you start out with an empty cache, the cache hit ratio
will be 0
>     until entries get into the cache and are pulled from the
cache rather
>     than the database.
>
>     Try doing a search like ldapsearch ... -b "basesuffixofdatabase"
>     "objectclass=*"
>     >
>
>
> Well, the servers are running for a long time, not only a days.
I have
> done that search, but the "Entry cache hit ratio" remains 0. I have
> also noticed that "Current entry cache size (in entries)" is only
> 4168, even after the search, although out directory cointains about
> 50000 entries. Is this normal?
We also recently fixed a bug with the cache size calculation.  What
platform?  What 389-ds-base version?  32-bit or 64-bit?
>

All servers are upgraded to 1.2.5 version, under CentOS 5.5 x86 (32 bit). Which is that bug? Is it in Bugzilla? \

Not sure, but I seem to recall this problem being fixed (perhaps as part of another bug fix) in 1.2.6.

---

-- 389 users mailing list 389-users@lists.fedoraproject.org https://admin.fedoraproject.org/mailman/listinfo/389-users