http://www.slony.info/bugzilla/show_bug.cgi?id=354

           Summary: failover can get stuck waiting for events after
                    sl_listen is empty
           Product: Slony-I
           Version: devel
          Platform: PC
        OS/Version: Linux
            Status: NEW
          Severity: enhancement
          Priority: low
         Component: stored procedures
        AssignedTo: slony1-bugs at lists.slony.info
        ReportedBy: ssinger at ca.afilias.info
                CC: slony1-bugs at lists.slony.info
   Estimated Hours: 0.0


This applies to the 2.2 branch and master

Consider a cluster where node 1 is the origin of all sets

1-->2
|
3->5
|
4

FAILOVER(id=5,backup node=5);

Where we are failing over a non-origin.

The remote worker for a node like node 4 might receiver an event from node 1

FAILOVER_NODE, 5, 5000001234)
That says the node 5 has failed and the last event from node 5 that node 1 has
seen is 5000001234

Now if node 4 gets the failover over before 5,5000001234
then the the remoteWorker_1 will call the failedNode() stored function on node
4.

Then it will wait for  5,5000001234.  However if slon then restarts the slon
will never process 5,5000001234 (a restart might not even be required but that
is how I've seen this)

The failedNode() call erased all the sl_listen entires with li_origin=5
select * FROM _disorder_replica.sl_listen where li_origin=5;
 li_origin | li_provider | li_receiver 
-----------+-------------+-------------
(0 rows)

None of the remoteListener threads will be listening for events from node 5 so
the misisng SYNC will never get processed.  The SYNC probably contains no data
because 5 is not an origin but the failover event is stuck.

-- 
Configure bugmail: http://www.slony.info/bugzilla/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are on the CC list for the bug.
You are the assignee for the bug.