Skip to content

The application periodically grinds to a near-halt (tasks that take seconds instead take 10+ minutes, users can't log in) for roughly an hour at a time, correlating with peak login/logout activity (around 9am, noon, and 4pm), with the Oracle alert log showing repeated ORA-3136 errors and one of the extended servers spiking to 100% memory during these episodes — the pattern is intermittent, not daily.

Answer

ORA-3136 specifically indicates a client failed to complete database authentication within the INBOUND_CONNECT_TIMEOUT window (60 seconds by default) — this is commonly caused by a client/application attempting to connect with an incorrect username or password, or by the database server being too heavily loaded to complete a client's logon within that timeout window even with correct credentials. Check the Middleware Server logs for correlating invalid-username/password errors around the same timestamps, and also look for 'Abandoning transaction after xx,xxx seconds' messages. Also check any additional installations that authenticate with IFS system accounts — Crystal Web Service, Print Agent, and similar — for stale or incorrect passwords for accounts like IFSPRINT/IFSSYS, since a misconfigured integration repeatedly retrying with bad credentials at a predictable interval can produce exactly this kind of periodic, load-correlated pattern. Taking thread dumps from the Middleware Servers during an active episode is also worth doing, given the CPU/memory spikes observed on extended servers. If this doesn't turn up a clear cause and you have Oracle support through IFS, opening a case referencing Oracle's own documentation on ORA-3136 inbound-connection-timeout troubleshooting is the next step, since your DBA may need Oracle-side assistance to fully resolve it.

← Browse Developer Studio, deployment, and upgrades questions