I have different ipa domains, installed on Red Hat 6.2 or CentOS 6.2
servers, with all 389-ds updates applied. After some time all the memory
in the server is consumed, mostly by the ns-slapd process.
I've looked at the RH Directory Server tuning manual at docs.redhat.com
and looked through the various tuning options. I found the
"nsslapd-cachememsize" and "nsslapd-dbcachesize" options to be set to
around 10MB by default. The id2entry.db4 file for the database in the
test IPA domain comes in at 6.7MB.
If I restart the directory server the memory is released. And then
slowly consumed again over time.
Running benchmarks with ldclt reveals much better numbers for the
directory servers that's recently been rebootet and does not yet consume
all the memory in the machine.
Am I missing some configuration? Is this normal behaviour for 389-ds to
consume all physical memory AND all swap memory? It certainly seem to
have a performance impact for the ldap server.
top - 21:31:51 up 52 days, 22:03, 2 users, load average: 0.33, 0.15, 0.07
Tasks: 136 total, 1 running, 135 sleeping, 0 stopped, 0 zombie
Cpu(s): 0.3%us, 0.7%sy, 0.0%ni, 92.7%id, 6.3%wa, 0.0%hi, 0.0%si,
Mem: 2055084k total, 1981892k used, 73192k free, 996k buffers
Swap: 1048568k total, 1048568k used, 0k free, 17172k cached
PID USER PR NI VIRT RES SHR S %CPU %MEM TIME+ COMMAND
1347 dirsrv 20 0 3301m 1.4g 2736 S 0.3 70.0 340:27.17 ns-slapd
Freeipa-users mailing list