Uploaded image for project: 'Confluence Data Center'
  1. Confluence Data Center
  2. CONFSERVER-17136

NonClusterManager does not prevent concurrent execution of jobs

      The implementation of ClusterManager for non clustered licenses does not implement any kind of locking at all.

      This means oncePerCluster (such as the ClusterSafetyJob) jobs can run concurrently, amongst other problems, which can lead to errors.

      Update: see the comment below for patch instructions.

        1. nonclustered.zip
          3 kB
          Partha
        2. NonClusterManager.class
          5 kB
          Andrew Lynch
        3. NonClusterManager.java
          4 kB
          Andrew Lynch
        4. NonClusterManager$SimpleClusteredLock.class
          1 kB
          Andrew Lynch

            [CONFSERVER-17136] NonClusterManager does not prevent concurrent execution of jobs

            Looks like we also ran into this when upgrading to 2.7.4. If that's likely, maybe adjusting the affected version list to include 2.7.4 might be a good idea?
            Before I upgraded to 2.7.4 I already read about this issue, but since it only stated 2.10.4 and 3.0.2 I assumed it doesn't apply to 2.7.4.

            Stefan Hett added a comment - Looks like we also ran into this when upgrading to 2.7.4. If that's likely, maybe adjusting the affected version list to include 2.7.4 might be a good idea? Before I upgraded to 2.7.4 I already read about this issue, but since it only stated 2.10.4 and 3.0.2 I assumed it doesn't apply to 2.7.4.

            FYI, that was the first of two bricked wikis for me in a three month period and I determined that Atlassian's software was too buggy to use. Sad, because I really like the company and the interface. But we installed it on a linux server; had this issue; reinstalled from scratch on a Windows server, upgraded the Sharepoint connector and ended up bricked again. I probably put in at least a week's work trying to recover from these crashes, on top of the standard installation and setup work, and I'm relieved that this all happened before we went into production. But I'm disappointed that we had to settle for Sharepoint's much poorer wikis. If Confluence is ever re-architected with back-end stability in mind, I'd be interested in cautiously reviewing the product again. But what I saw was a product that isn't out of alpha on the back-end side, despite having a highly sophisticated and elegant UI.


            Peter

            Peter Campbell added a comment - FYI, that was the first of two bricked wikis for me in a three month period and I determined that Atlassian's software was too buggy to use. Sad, because I really like the company and the interface. But we installed it on a linux server; had this issue; reinstalled from scratch on a Windows server, upgraded the Sharepoint connector and ended up bricked again. I probably put in at least a week's work trying to recover from these crashes, on top of the standard installation and setup work, and I'm relieved that this all happened before we went into production. But I'm disappointed that we had to settle for Sharepoint's much poorer wikis. If Confluence is ever re-architected with back-end stability in mind, I'd be interested in cautiously reviewing the product again. But what I saw was a product that isn't out of alpha on the back-end side, despite having a highly sophisticated and elegant UI. — Peter

            scayla added a comment -

            I came across quite the same error as Peter Campbell. Cleaning the plugins-temp directory seems to resolve the fatal issue.

            scayla added a comment - I came across quite the same error as Peter Campbell. Cleaning the plugins-temp directory seems to resolve the fatal issue.

            Source code attached.

            Andrew Lynch (Inactive) added a comment - Source code attached.

            Linh Pham added a comment -

            Andrew - before I do the patch, I would like to review the new code. could you email the code for the 2 classes? Thx. linh at usc dot edu.

            Linh Pham added a comment - Andrew - before I do the patch, I would like to review the new code. could you email the code for the 2 classes? Thx. linh at usc dot edu.

            Hi Cyrille,

            You should raise a support request at https://support.atlassian.com if this is the case.

            Regards,
            Andrew Lynch

            Andrew Lynch (Inactive) added a comment - Hi Cyrille, You should raise a support request at https://support.atlassian.com if this is the case. Regards, Andrew Lynch

            Hi,

            The problem is still here on a Confluence 3.0.2 with the addition of class files in WEB-INF/classes/com/atlassian/confluence/cluster/nonclustered and a shutdown / restart.

            Does anyone have the same behaviour ??

            Cyrille Martin added a comment - Hi, The problem is still here on a Confluence 3.0.2 with the addition of class files in WEB-INF/classes/com/atlassian/confluence/cluster/nonclustered and a shutdown / restart. Does anyone have the same behaviour ??

            jens added a comment -

            We are sincerely sorry for not updating our release notes earlier. We have learned from this mistake and will handle critical issues like this more appropriately in the future.

            jens added a comment - We are sincerely sorry for not updating our release notes earlier. We have learned from this mistake and will handle critical issues like this more appropriately in the future.

            Partha added a comment -

            The patch provided will work for 2.10.4.

            To use it, please unzip the attached zip into WEB-INF/classes/com/atlassian/confluence/cluster/nonclustered (you will need to create the last couple of directories) and restart your Confluence instance.

            Kind Regards,
            Partha Kamal

            Partha added a comment - The patch provided will work for 2.10.4. To use it, please unzip the attached zip into WEB-INF/classes/com/atlassian/confluence/cluster/nonclustered (you will need to create the last couple of directories) and restart your Confluence instance. Kind Regards, Partha Kamal

            Mark Nye added a comment -

            We were just bitten by this one as well and it's extremely frustrating. We've got several thousand people who depend on our instance to get their work done, including instructors who teach classes on Confluence. Here's my question: Why bother fixing critical bugs in recent versions if don't intend to release an updated version (3.0.3) or even document the problem?

            Atlassian's recruiting material goes on and on about how smart and cool everyone is, but folks, setting up these kind of traps for your enterprise customers is neither smart nor cool. It is however a great way to get us to start looking at other collaboration products.

            Mark Nye added a comment - We were just bitten by this one as well and it's extremely frustrating. We've got several thousand people who depend on our instance to get their work done, including instructors who teach classes on Confluence. Here's my question: Why bother fixing critical bugs in recent versions if don't intend to release an updated version (3.0.3) or even document the problem? Atlassian's recruiting material goes on and on about how smart and cool everyone is, but folks, setting up these kind of traps for your enterprise customers is neither smart nor cool. It is however a great way to get us to start looking at other collaboration products.

              alynch Andrew Lynch (Inactive)
              alynch Andrew Lynch (Inactive)
              Affected customers:
              1 This affects my team
              Watchers:
              11 Start watching this issue

                Created:
                Updated:
                Resolved: