You are viewing a plain text version of this content. The canonical link for it is here.
Posted to issues@sentry.apache.org by "Alexander Kolbasov (JIRA)" <ji...@apache.org> on 2016/10/12 22:09:20 UTC
[jira] [Comment Edited] (SENTRY-1491) Sentry transactions are not
rolled back immediately when commit fails
[ https://issues.apache.org/jira/browse/SENTRY-1491?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15570018#comment-15570018 ]
Alexander Kolbasov edited comment on SENTRY-1491 at 10/12/16 10:08 PM:
-----------------------------------------------------------------------
I think that we should close persistance manager using try-with-resources which requires PersistentManager to be closeable. This requires SENTRY-1500.
was (Author: akolb):
I think that we should close transactions using try-with-resources which requires PersistentManager to be closeable. This requires SENTRY-1500.
> Sentry transactions are not rolled back immediately when commit fails
> ---------------------------------------------------------------------
>
> Key: SENTRY-1491
> URL: https://issues.apache.org/jira/browse/SENTRY-1491
> Project: Sentry
> Issue Type: Bug
> Components: Sentry
> Affects Versions: 1.7.0, sentry-ha-redesign
> Reporter: Alexander Kolbasov
> Assignee: Alexander Kolbasov
>
> Here is the code for commitTransaction:
> {code}
> public void commitTransaction(PersistenceManager pm) {
> Transaction currentTransaction = pm.currentTransaction();
> try {
> Preconditions.checkState(currentTransaction.isActive(), "Transaction is not active");
> currentTransaction.commit();
> } finally {
> pm.close();
> }
> }
> {code}
> so it does pm.close() in the finally clause, so it always closes pm.
> Now everyone else does something like
> {code}
> commitTransaction(pm);
> rollbackTransaction = false;
> return roles;
> } finally {
> if (rollbackTransaction) {
> rollbackTransaction(pm);
> }
> }
> {code}
> which means that pm is closed by the time we call rollbackTransaction() and we never rollback when commit fails. As a result rollback is deferred till connection close (which may be some time due to connection pooling) and we are potentially holding various locks for longer than needed.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)