Red Hat Bugzilla – Bug 1316498
OData - if PUT fails, next update request fails because table in DB is still locked
Last modified: 2016-08-24 07:42:55 EDT
Description of problem:
I can see (from the log) that processing of PUT request is as follows:
1. select record from table
2. delete record
3. parse entity from HTTP request
4. insert new record
But if e.g. step 3 fails, Teiid cancels transaction and returns error. In my opinion, parsing of the entity should be the first step (or at least right after select).
What happen here is, that DELETE statement creates transaction and locks the table. But the transaction is not canceled immediately after cancel request . This causes that no other query can succeed, because of lock on the table (EAP uses connection pool and does not close connections immediately and therefore open connection has still one transaction in active state).
*Note:* We use H2 database for testing purpose. But I think that this could happen with any DB. Anyway, parsing of the entity should not be performed after delete.
09:35:33,937 DEBUG [org.teiid.TRANSPORT] (New I/O worker #1) Channel closed
09:40:31,430 WARN [com.arjuna.ats.arjuna] (Transaction Reaper) ARJUNA012117: TransactionReaper::check timeout for TX 0:ffff7f000001:113c3cbe:56e131a7:e in state RUN
09:40:31,438 WARN [com.arjuna.ats.arjuna] (Transaction Reaper Worker 0) ARJUNA012121: TransactionReaper::doCancellations worker Thread[Transaction Reaper Worker 0,5,main] successfully canceled TX 0:ffff7f000001:113c3cbe:56e131a7:e
Steven Hawkins <email@example.com> updated the status of jira TEIID-4057 to Coding In Progress
Steven Hawkins <firstname.lastname@example.org> updated the status of jira TEIID-4057 to Resolved
Steven Hawkins <email@example.com> updated the status of jira TEIID-4057 to Closed