Thanks for the feedback msymons!

Hi francois.nonnenmacher, that's correct. Fix version is 5.8.6.

Except that it isn't fixed on 5.8.x. On Confluence 5.8.5 in production, running Hercules will generate pages and pages of warnings:

WARN [hercules] [atlassian.confluence.cache.TransactionalCacheFactory] logNonTxUsageWarning Update operation performed on transactional cache [stp.task.monitor.hercules] outside of a transaction. All updates to this cache should be performed from a thread with a valid transaction context.

jens.kasperek Just to give you some background.

Non-transactional access to transactional caches in confluence was from version 1.0. In version 5.x (I don't recall whenever that was exactly 5.2.5) we added this warning message to track these caches down and we pretty much fixed all confluence codebase in v 5.7.

So why message still appears??? The reason are cross product plugins (example: Support Tools Plugin) and plugin based scheduled jobs which use cache API shared between JIRA/Confluence/Stash. STP plugin doesn't need transactional cache, however the way we expose cache factory -> it could get only transactional cache therefore you are getting warning in log files. In JIRA, for example there is no concept of transactional caches at all (it's Confluence only thing) so if value was put into cache during transaction and transaction failed then this entry will be poisoned in JIRA (in confluence you'll get cache changes won't be committed into main cache).

While confluence specific caches have been fixed, we can't easy update cross product plugins/scheduled jobs (that's yet another topic).

What has been done? We logging this message in mode that it visible only to confluence developers and on our test instance. Warning won't be visible on production and won't be cluttering logs. It's not exactly "fix", rather band aid, but it's the best I can do at the moment.

Now as 5.8.6 (5.8.x series) are the fastest version of confluence we released in last few years. I have right to assume that this issue is watched by sysadmins who watching errors in logs at daily basis and have large number of users to care about. I really recommend to upgrade to 5.8.x. I do guarantee large of audience of happy people commenting on confluence post upgrade

I was just informed, that this issue is resolved with Conf 5.8.6 or Conf 5.9. Can we know what resolves this and if we can implement the fix in a prior version?

I confirm the same bug with latest Confluence 5.8.5, it seems to appear a LOT when doing a support request:

 -- referer: https://xxx/plugins/servlet/stp/view/ | url: /plugins/servlet/stp/view/create-support-request/execute | userName: sorins
2015-07-01 17:18:10,884 WARN [support-request] [atlassian.confluence.cache.TransactionalCacheFactory] logNonTxUsageWarning Update operation performed on transactional cache [stp.task.monitor.support-request] outside of a transaction. All updates to this cache should be performed from a thread with a valid transaction context.
 -- referer: https://xxx/plugins/servlet/stp/view/ | url: /plugins/servlet/stp/view/create-support-request/execute | userName: sorins

Here are two examples:

2015-06-23 07:27:17,541 WARN [http-bio-8080-exec-85] [confluence.util.profiling.DurationThresholdWarningTimingHelperFactory] logMessage Execution time for publishing event com.atlassian.confluence.event.events.content.page.PageMoveCompletedEvent[source=page: Collaboration Model v.16 (293886071)] took 12331 ms (warning threshold is 5000 ms)
 -- url: /confluence/pages/movepage.action | userName: nac5kor | referer: https://oururl.com/confluence/display/cmmtg/Collaboration+Model | action: movepage
2015-06-23 07:27:20,811 WARN [http-bio-8080-exec-85] [atlassian.confluence.cache.TransactionalCacheFactoryCleanupFilter] checkBoundTransactionalCaches Found some transactional caches still bound to thread following execution of [/confluence/pages/movepage.action] - [com.atlassian.bandana.BandanaPersister]
 -- url: /confluence/pages/movepage.action | userName: nac5kor | referer: https://oururl.com/confluence/display/cmmtg/Collaboration+Model
2015-06-23 07:27:22,974 WARN [http-bio-8080-exec-85] [atlassian.confluence.cache.TransactionalCacheFactoryCleanupFilter] checkAndForceUnbindOfTransactionalCaches Forcibly unbound thread-local transactional caches prior to request execution: [com.atlassian.bandana.BandanaPersister]
 -- url: /confluence/rest/rb-core-change/1.0/isCurrentUserConfluenceAdmin | userName: ACH6FE | referer: https://oururl.com/confluence/pages/templates2/listpagetemplates.action?key=leandobd
2015-06-23 07:27:22,975 WARN [http-bio-8080-exec-85] [atlassian.confluence.cache.TransactionalCacheFactoryCleanupFilter] checkAndForceUnbindOfTransactionalCaches Forcibly unbound thread-local transactional cached references prior to request execution: []
 -- url: /confluence/rest/rb-core-change/1.0/isCurrentUserConfluenceAdmin | userName: ACH6FE | referer: https://oururl.com/confluence/pages/templates2/listpagetemplates.action?key=leandobd
2015-06-23 07:27:30,015 WARN [scheduler_Worker-7] [atlassian.confluence.cache.TransactionalCacheFactory] warning Transactional cache update outside transaction. All updates to this cache should be performed from a thread with a valid transaction context.
2015-06-23 07:27:30,017 WARN [scheduler_Worker-7] [atlassian.confluence.cache.TransactionalCacheFactory] warning Transactional cache update outside transaction. All updates to this cache should be performed from a thread with a valid transaction context.
2015-06-23 07:27:30,020 WARN [scheduler_Worker-7] [atlassian.confluence.cache.TransactionalCacheFactory] warning Transactional cache update outside transaction. All updates to this cache should be performed from a thread with a valid transaction context.
2015-06-23 07:27:30,022 WARN [scheduler_Worker-7] [atlassian.confluence.cache.TransactionalCacheFactory] warning Transactional cache update outside transaction. All updates to this cache should be performed from a thread with a valid transaction context.
2015-06-23 07:27:30,024 WARN [scheduler_Worker-7] [atlassian.confluence.cache.TransactionalCacheFactory] warning Transactional cache update outside transaction. All updates to this cache should be performed from a thread with a valid transaction context.
..

and

2015-06-24 09:49:37,596 WARN [http-bio-8080-exec-150] [persistence.dao.hibernate.HibernateDraftDao] getDraft Found 3 drafts for pageId = 0 creator = 'com.atlassian.confluence.user.ConfluenceUserImpl@334feb96[olc2rt,8a83a02c4a10e5f1014a11aa51fa7faa]' draftType = 'page'
 -- url: /confluence/pages/createpage.action | userName: olc2rt | referer: https://oururl.bosch.com/confluence/display/PHOT/Applications | action: createpage
2015-06-24 09:51:00,025 WARN [scheduler_Worker-2] [atlassian.confluence.cache.TransactionalCacheFactory] warning Transactional cache update outside transaction. All updates to this cache should be performed from a thread with a valid transaction context.
2015-06-24 09:52:09,420 WARN [http-bio-8080-exec-87] [auth.trustedapps.filter.TrustedApplicationFilterAuthenticator] authenticate Failed to login trusted application [jira:14051493] due to bad URL signature. Received protocol version [2]. Required protocol version [2]

You see, sometimes we might have a relation to a user action. Sometimes, these log statements just occur without any further hint.

msymons ok, thanks. Reopening issue.

In the understanding that the warnings are being cleaned up one by one, here are a couple from from logs for Confluence v5.7.4:

2015-06-24 01:37:42,859 INFO [scheduler_Worker-4] [atlassian.crowd.directory.DbCachingRemoteDirectory] synchroniseCache INCREMENTAL synchronisation complete for directory [ 12877825 ] in [ 98ms ]
2015-06-24 02:10:00,106 WARN [scheduler_Worker-2] [atlassian.confluence.cache.TransactionalCacheFactory] logNonTxUsageWarning Update operation performed on transactional cache [stp.task.monitor.hercules] outside of a transaction. All updates to this cache should be performed from a thread with a valid transaction context.
2015-06-24 02:10:00,110 WARN [scheduler_Worker-2] [atlassian.confluence.cache.TransactionalCacheFactory] logNonTxUsageWarning Update operation performed on transactional cache [stp.task.monitor.hercules] outside of a transaction. All updates to this cache should be performed from a thread with a valid transaction context.
2015-06-24 02:10:03,858 WARN [hercules] [atlassian.confluence.cache.TransactionalCacheFactory] logNonTxUsageWarning Update operation performed on transactional cache [stp.task.monitor.hercules] outside of a transaction. All updates to this cache should be performed from a thread with a valid transaction context.
2015-06-24 02:37:42,756 INFO [scheduler_Worker-5] [atlassian.crowd.directory.DbCachingRemoteDirectory] synchroniseCache INCREMENTAL synchronisation for directory [ 12877825 ] starting
2015-06-24 02:37:42,760 INFO [scheduler_Worker-5] [atlassian.crowd.directory.DbCachingRemoteDirectory] synchroniseCache Attempting INCREMENTAL synchronisation for directory [ 12877825 ]

I have included an hour's worth of log entries so that you can see the transactions cache warnings in context.

jens.kasperek would you be able to attach some snippets of your logs files to issue.

We are running Confluence on version 5.5.6 and we also get hundreds of these messages per day on our log files:
2015-06-22 10:11:30,084 WARN [scheduler_Worker-6] [atlassian.confluence.cache.TransactionalCacheFactory] warning Transactional cache update outside transaction. All updates to this cache should be performed from a thread with a valid transaction context.

So, this issue does not seem to be fixed since 5.5.3!?! Please re-open this issue!

heimi workaround is to flush `Page Children` cache on admins screen.

Url on your instance would be `/admin/cache/showStatistics.action#fullView`.

Sorry about this bug, we are working towards better transaction isolation atm.

I have these warnings too (on 5.5.2) but there is also broken functionality. The warning appears every time a user moves a page in the hirarchical view. The page is first moved in the hirarchical view, but when you return to the Wiki, it is on the former position. Any known workaround for that ?

msymons, the warnings haven't been eliminated entirely, there's still a few of them hanging around. As you said, they're generally associated with specific functions now. The original bug as reported was that these warnings were being reported continuously and for pretty much every function, due to the transactional cache mechanism being accidentally broken in those versions. The remaining warnings in 5.5.6 are valid, but the functionality appears to be unaffected. We hope to address all of the remaining examples over the next few months (sooner if one of them proves to be causing real problems).

Are you sure that this issue is fixed? I am still seeing it with Confluence v5.5.4

However, there is a context to the events... they seem to be connected to a wiki import:

2014-08-19 14:53:58,754 INFO [http-bio-8090-exec-2250] [confluence.rpc.auth.TokenAuthenticationInvocationHandler] invokeAuthenticatedMethod Executing remote API method [ movePage ] as user [ userx ]
2014-08-19 14:53:59,823 INFO [http-bio-8090-exec-2256] [confluence.rpc.auth.TokenAuthenticationInvocationHandler] invokeAuthenticatedMethod Executing remote API method [ convertWikiToStorageFormat ] as user [ userx ]
2014-08-19 14:54:00,291 WARN [scheduler_Worker-8] [atlassian.confluence.cache.TransactionalCacheFactory] warning Transactional cache update outside transaction. All updates to this cache should be performed from a thread with a valid transaction context.
2014-08-19 14:54:00,332 WARN [scheduler_Worker-8] [atlassian.confluence.cache.TransactionalCacheFactory] warning Transactional cache update outside transaction. All updates to this cache should be performed from a thread with a valid transaction context.
2014-08-19 14:54:00,702 INFO [http-bio-8090-exec-2235] [confluence.rpc.auth.TokenAuthenticationInvocationHandler] invokeAuthenticatedMethod Executing remote API method [ getPermissions ] as user [ userx ]
2014-08-19 14:54:01,098 INFO [http-bio-8090-exec-2255] [confluence.rpc.auth.TokenAuthenticationInvocationHandler] invokeAuthenticatedMethod Executing remote API method [ getSpace ] as user [ userx ]

When performing Content Indexing, this message is written to the log many times per second. Using Confluence 5.5.2.

No obvious ill-effects – other than rapid log-growth – so far, however.

We are seeing this in 5.5.2 version.

We have this same behavior in Confluence 5.2.5

Mikael, I have experienced the same thing. It is hard to reproduce because sometimes it doesn't save anything, and then other times it will not. There is no real smoking gun in the logs.

I can confirm the finding by Iris:
We upgraded to Confluence 5.5.2 yesterday, and now Confluence do not save a draft of the page I am editing. The draft save interval is set to 30 seconds.
I edit the page, do some changes, wait a few minutes, kill the browser, start the browser, go back to the page and click edit.
Expected result: A notification about a saved draft version, and some options to retrieve or discard it.
Actual result: The page editor just come up with the original content of the page.

EDIT: I have been unable to reproduce this error, and so has my colleague. I suspect I made some sort of mistake in my initial investigation which led me to believe there was an error with saving drafts.
Sorry about the confusion.

I am seeing this in 5.5.2. I think it is affecting drafts being saved. The pages it mentions are pages I am messing with to see if drafts are saving, which they area not.

mcobby, 5.5.2 should be fit for purpose, this problem is highly unlikely to affect production systems. The warning in the logs is really for the Confluence developers to spot potential problems, rather than alerting about actual problems. However, 5.5.3 is currently in the process of being built, so if you have any doubts then you may want to wait for that. The problem has been fully resolved in that version.

Can you please confirm if 5.5.2 is fit for production use or not? I refer to CONF-33774

While this change did go into stable, at this point we do not believe that it is likely to have a noticeable impact on production systems. If it does, it'll most likely have similar impacts to the test failure messages mentioned above, where caches related to users or user directories go stale.

It doesn't sound like it's a trivial error for systems with 1000s of users. About to push through an upgrade to 5.5.2 and I would like to know if I'm need to pull the project.

This is not a minor bug, this could put a heavely usage instance down by filling the logs!

Same problem after upgrade to 5.5

I'm seeing the same problem on 5.5.2

Same here. Just upgraded from 5.1 to 5.5.1

Same here, just upgraded from 5.4.4 to 5.5.1. Also saw a continuous scroll of these when I reindexed.

Same here. I just upgraded from 5.4.4 to 5.5.1 and I'm seeing this in the log every time I click on a link.

I am also seeing a similar problem on 5.5.1