Hi,
I made some tests with the 2 solutions:
1. checkpoint directive in cron job (my reference about this solution:
http://www.openldap.org/lists/openldap-bugs/200306/msg00096.html)
2. checkpoint directive in slapd.conf "checkpoint 10000 60" according to
your advice
My test was the same for the 2 solutions : mass-provisioning of 1000 new
subscriptions (1 subscription = 4 ldap entry in bdb1 backend + 3 ldap
entry in bdb2 backend => cf. slapd.conf for backends definition)
Here's my results:
solution 1:
time to execute the provisioning = 6mn43s
only 2 running slapd with %CPU = 1-2% in average
db_stat with option -e, -m, -t, -c, -l in the 2 backend seems to be
correct
solution2:
time to execute the provisioning = 20mn28s
8 running slapd with %CPU = 20% in average (with for one of them a peak =
32%CPU)
db_stat with options -m, -t, -c in the 2 backend seems to be correct
db_stat with option -l gives a bad result: "1355 The number of region
locks granted after waiting." for bdb1 backend
and "3411 The number of region locks granted after waiting." for bdb2
backend
Cf. in attachment the result of db_stat command for solution n°2 (only
bdb1 backend)
Cf. in attachment too the DB_CONFIG file of the bdb1 backend for example
So my question is: How doing to obtain best results with solution n°2
(with tuning BerkeleyDB ??)
the solution n°1 being disadvised ?