locks persists while client app is down

We are running IDS 12.1 FC6 on Solaris. Last day a unix box residing java applications gone irresponsive due to high load on it. We restarted the machine but there were locks even the applications were down (these applications maintain jdbc connection pool with IDS database). We think that transactions are rolled back and sessions terminate if we application is shutdown, but now we are in doubt. If sessions persists, can we figure out which sessions are in orphan state?

Assuming this situation no longer exists .... If you're sure these were left-over application locks, than I'd assume 'global transactions' - perfectly possible to have these persist, incl. their locks, beyond the life of an application. From: "KAMRAN HAQ" <khaq@i2cinc.com> To: ids@iiug.org Date: 10/11/2017 04:54 PM Subject: locks persists while client app is down [40052] Sent by: ids-bounces@iiug.org We are running IDS 12.1 FC6 on Solaris. Last day a unix box residing java applications gone irresponsive due to high load on it. We restarted the machine but there were locks even the applications were down (these applications maintain jdbc connection pool with IDS database). We think that transactions are rolled back and sessions terminate if we application is shutdown, but now we are in doubt. If sessions persists, can we figure out which sessions are in orphan state? ******************************************************************************* Forum Note: Use "Reply" to post a response in the discussion forum.

Do the java applications make use of xa transactions ? Check if you have global transaction IDs: onstat -G Then check for lock with: onstat -k and search for locks with the owner "address" between parentheses (it means the lock is held by a xa transaction). You can also search for transactions with the "userthread" equal to zero (which usually means that it is a xa transaction): onstat -x Luis Marques -----Original Message----- From: ids-bounces@iiug.org [mailto:ids-bounces@iiug.org] On Behalf Of KAMRAN HAQ Sent: quarta-feira, 11 de outubro de 2017 15:54 To: ids@iiug.org Subject: locks persists while client app is down [40052] We are running IDS 12.1 FC6 on Solaris. Last day a unix box residing java applications gone irresponsive due to high load on it. We restarted the machine but there were locks even the applications were down (these applications maintain jdbc connection pool with IDS database). We think that transactions are rolled back and sessions terminate if we application is shutdown, but now we are in doubt. If sessions persists, can we figure out which sessions are in orphan state? ******************************************************************************* Forum Note: Use "Reply" to post a response in the discussion forum.

We do not explicitly using xa transactions, but database instance has HDR and 2 RSS connected in near to sync mode. May be that cluster is the reason? Can we terminate the sessions we identifies as global xa if we could, to avoid instance restart?

If you confirm that the locks belong to Global Transactions (aka xa transactions), then there is no direct way to clear them without a restart (and they will probably survive the restart??). One way is checking the RFE in: http://www.ibm.com/developerworks/rfe/execute?use_case=viewRfe&CR_ID=36910 The RFE has an example of an ESQLC program that can remove Global Transactions. Another way is to force a transaction rollback due to a long transaction. Change the value of LTXHWM to a very low value and force changes of logical logs with "onmode -l". BE CAREFUL if you try this, because you may force the rollback of other transactions.

Only a shot in the dark: What about onmode -Z. May Kill a distributed transaction. (https://www.ibm.com/support/knowledgecenter/en/SSGU8G_12.1.0/com.ibm.adref.doc/ ids_adr_0443.htm) -----Ursprüngliche Nachricht----- Von: ids-bounces@iiug.org [mailto:ids-bounces@iiug.org] Im Auftrag von KAMRAN HAQ Gesendet: Freitag, 13. Oktober 2017 13:49 An: ids@iiug.org Betreff: Re: RE: locks persists while client app is down [40064] We do not explicitly using xa transactions, but database instance has HDR and 2 RSS connected in near to sync mode. May be that cluster is the reason? Can we terminate the sessions we identifies as global xa if we could, to avoid instance restart? ******************************************************************************* Forum Note: Use "Reply" to post a response in the discussion forum.

Thank you all, Hopefully I am prepared to tackle the issue if it happens again. Mean while I try to reproduce the it and see how the suggested methods help to diagnose and fix.