RECO Process Not Clearing Indoubt Transactions After A Database Switchover

刘泽宇

现象：
The Recover(RECO) process for handling in-doubt distributed transactions does not work  after a database switchover in a database(Involving Standby configuration) which acts as a XA resource and participates in distributed transactions.

The problem is that the entries in the table DBA_2PC_PENDING never get recovered or purged.

The column RETRY_TIME in the table DBA_2PC_PENDING will not be updated as well. It will remain the  same as the timestamp in  the FAIL_TIME column.

In normal cases the database will do a retry of the transactions ,when a distributed transactions fails into DBA_2PC_PENDING.

原因：
This happens in any environment including standalone or multi-tenant databases where a standby database configuration exists.

The problem was identified as Bug 28298401 which has been closed as a duplicate of Bug 34367505.

处理方法：
To resolve this issue,please download and apply patch 34367505 .The patch is backport feasible.

This Bug is fixed in database Version->20.1

可行措施:

As a workaround,One can perform the below actions,

1) Post-Switchover , restart the database which will make the RECO to function normally and clear new in-doubt transactions (if any).

2) Kill the RECO process after switchover. In its next spawn,it will clean new in-doubt transactions (if any).

NOTE : If a patch for your version and platform is not available, please contact Global Customer Services through a Service Request and attach output of opatch lsinventory and uname -a, in order to request a backport