[Building Sakai] LDAP receives bind and abandon requests under high load

[Stephen Jaegle]

Hi all:

After going into production this Summer at URI with Sakai and LDAP, we have
just started Fall with vastly higher concurrent sessions and critical
performance issues relating to LDAP.

Our LDAP logs eventually reach a state where one sees only binds followed by
immediate abandon requests occurring within the same second, with no
searches taking place, so that users are unable to log in because they
cannot be authenticated. Restarting tomcat seems to relieve the problem for
awhile, on the order of < 1hr; leaving tomcat up and stopping the LDAP
server for a few minutes and then restarting it also seems to relieve the
issue for a similar period of time, though simply stopping and restarting
LDAP with no delay does not change the condition at all.

Yesterday, the logs showed about 300 abandons, while today they showed over
11,000 abandons with only about twice the number of concurrent sessions.
Rebooting the server that our two load-balanced tomcat instances are running
on seems to have drastically reduced the issue as virtually no abandons have
occurred since that time, even though the number of concurrent sessions has
risen to levels comparable to earlier today when LDAP logs showed only a
series of bind-abandon sequences.

We are currently running Sakai 2.5.4 with:
operationTimeout = 5000
cacheTTL = 300000
pooling = false

Any help would be appreciated.

Thanks,

Steve
-- 
View this message in context: http://www.nabble.com/LDAP-receives-bind-and-abandon-requests-under-high-load-tp25374261p25374261.html
Sent from the Sakai - Development mailing list archive at Nabble.com.