active session waiting for SQL*Net more data from client for very long time
From: Denis <denis.sun_at_yahoo.com>
Date: Fri, 20 Jan 2012 07:27:48 -0800 (PST)
Message-ID: <1327073268.38266.YahooMailNeo_at_web161804.mail.bf1.yahoo.com>
Hi, Listers,
Recently we have encountered sessions ( from weblogic JDBC thin client) that are waiting for "SQL*Net more data from client" event and in active status indefinitively (could be more than 10 hours). In two instances, such sessions eventually block other sessions in the morning when business transactions become busy. we have to kill such a session to release the locks it held to avoid serious lost of service.
The sessions are doing a relatively simple select statement with very selective index scan when they were hanging with above event. The largest rows it can return at most several hundred. By google, there is a case described here: http://blogs.warwick.ac.uk/java/entry/wait_class_network/ , which solved this issue by upgrading JDBC driver. But I am not sure this is reason for my case as we use same JDBC driver for all app server and we don't have long sql text as described in the webpage. ( to be honest I have little knowlege about weblogic that managed by other team and to get other team's coordination requires a bit more effort).
Our dev team is trying to enable time-out at statement level in their java code. Hope this can prevent such kind of running away transaction.(Weblogic transaction time-out seems not working as they said they set the value for 180s already) Anyone has similar experience and can give me some clue to dig into the root cause?
Thanks,
Denis
Date: Fri, 20 Jan 2012 07:27:48 -0800 (PST)
Message-ID: <1327073268.38266.YahooMailNeo_at_web161804.mail.bf1.yahoo.com>
Hi, Listers,
Recently we have encountered sessions ( from weblogic JDBC thin client) that are waiting for "SQL*Net more data from client" event and in active status indefinitively (could be more than 10 hours). In two instances, such sessions eventually block other sessions in the morning when business transactions become busy. we have to kill such a session to release the locks it held to avoid serious lost of service.
The sessions are doing a relatively simple select statement with very selective index scan when they were hanging with above event. The largest rows it can return at most several hundred. By google, there is a case described here: http://blogs.warwick.ac.uk/java/entry/wait_class_network/ , which solved this issue by upgrading JDBC driver. But I am not sure this is reason for my case as we use same JDBC driver for all app server and we don't have long sql text as described in the webpage. ( to be honest I have little knowlege about weblogic that managed by other team and to get other team's coordination requires a bit more effort).
Our dev team is trying to enable time-out at statement level in their java code. Hope this can prevent such kind of running away transaction.(Weblogic transaction time-out seems not working as they said they set the value for 180s already) Anyone has similar experience and can give me some clue to dig into the root cause?
Thanks,
Denis
-- http://www.freelists.org/webpage/oracle-lReceived on Fri Jan 20 2012 - 09:27:48 CST