1 2 Previous Next 18 Replies Latest reply on Feb 6, 2012 11:03 PM by skidvd Go to original post
      • 15. Re: JMS Message expiration with paging
        skidvd

        So I ran two additional tests at Cleberts's request.

         

        In the first test, I introduced another full/clean restart of both live and backup servers as well as the remaining connected clients.  So, in short I initially brought up both servers and both clients.  No pubs.  I then shut everything down and restarted both severs and ony one of the clients - leaving one durable subscription configured by with no active consumer on it.  Started pubs and then got to PAGE level.  All expiration were just fine until I failed the live server.  After failover, no further expiration out of the page file ever occurred from that point on.  Works great before any cluster failure - does not work (expire) after any cluster failure.

         

        In the second test, I started both the live and backup servers, but had no consumers at all.  I then started the publication (again no consumers per Clebert's request).  As expected, we never reached the PAGEing condition as these are durable topics - without a durable subscription, there would be no reason to ever reach a PAGEing condition with no consumers.

         

        There is clearly something related to cluster failure that prevents the ongoing operation of expiration of messages out of the PAGE files. 

        • 16. Re: JMS Message expiration with paging
          clebert.suconic
          • 17. Re: JMS Message expiration with paging
            clebert.suconic

            I already committed it on AS7 and EAP branch. I'm going to move it to trunk (and I will then close the JIRA). But it should be solved on AS7 branch already.

             

             

            It was a minor issue. I almost wasn't going to fix it.. as you would just need a consumer actve to kick depaging, and after that everything would work fine.

            • 18. Re: JMS Message expiration with paging
              skidvd

              Clebert Suconic wrote:

               

              I already committed it on AS7 and EAP branch. I'm going to move it to trunk (and I will then close the JIRA). But it should be solved on AS7 branch already.

               

               

              It was a minor issue. I almost wasn't going to fix it.. as you would just need a consumer actve to kick depaging, and after that everything would work fine.

               

              The importance and crtiticality of this issue depends on the environment and consumers.  Have an active consumer to solve this problem is only true if you are sure your consumers are both known and well-behaved.  As in the test scenario, any single remaining consumer is NOT sufficient to kick depaging.  In our environment, we often have transient consumers that come and go - and occasionally NEVER return.  If this were not fixed, we would have no way of creating a consumer and certainly have not guarantee that they would ever return - therefore causing likely disastrous endless paging growth.  We use expiration to guard ourselves from this, but need to know that the system will reliably act on that expiration for all potential circumstances including failover.

              1 2 Previous Next