CAS 5.1.9 DefaultTicketRegistryCleaner fails when TGT not found

448 views
Skip to first unread message

Juan Quintanilla

unread,
Mar 19, 2019, 11:51:21 AM3/19/19
to cas-...@apereo.org

Hello,


We are running 2 CAS 5.1.9 instances with ORacle 12 for the JPA ticket registry and we seeing the following error in the logs:


ERROR [org.apereo.cas.ticket.registry.DefaultTicketRegistryCleaner] - <Unable to find org.apereo.cas.ticket.TicketGrantingTicketImpl with id TGT 


The ticket mentioned in the logs belongs to the 2nd webserver and the cleanup process seems to be only running on the 1st webserver every 30 minutes even though it configured on both.  It never goes to the second webserver. The problem we run into is that because the cleanup process fails to find the ticket everything is rolled back and no ticket is removed.  From what we can find the issue occurs when the TGT is removed from the 2nd server because of a user initiated logout at the same time that the 1st server initiates the cleanup process.  


Has anyone encountered something similar is there maybe a workaround to this so that tickets are cleaned up. Below is the configuration we have for the Ticket Registry and Cleanup Process:


cas.ticket.registry.jpa.ticketLockType=NONE
cas.ticket.registry.jpa.jpaLockingTimeout=3600

#cas.ticket.registry.jpa.healthQuery=select 1
cas.ticket.registry.jpa.isolateInternalQueries=false
cas.ticket.registry.jpa.url=xxxxxxxxx
cas.ticket.registry.jpa.failFast=true
cas.ticket.registry.jpa.dialect=org.hibernate.dialect.Oracle12cDialect
cas.ticket.registry.jpa.leakThreshold=10
#cas.ticket.registry.jpa.jpaLockingTgtEnabled=false
cas.ticket.registry.jpa.batchSize=10
#cas.ticket.registry.jpa.defaultCatalog=
#cas.ticket.registry.jpa.defaultSchema=
cas.ticket.registry.jpa.user=xxxxxxxxxx
cas.ticket.registry.jpa.ddlAuto=update
cas.ticket.registry.jpa.password=xxxxxxxxxxx
cas.ticket.registry.jpa.autocommit=false
cas.ticket.registry.jpa.driverClass=oracle.jdbc.OracleDriver
cas.ticket.registry.jpa.idleTimeout=5000
#cas.ticket.registry.jpa.dataSourceName=
cas.ticket.registry.jpa.dataSourceProxy=false
cas.ticket.registry.jpa.pool.suspension=false



#Added to help with JPA Deadlock, space out the ticket cleanup
# cas.ticket.registry.cleaner.appId=cas-ticket-registry-cleaner
# cas.ticket.registry.cleaner.startDelay=10000
cas.ticket.registry.cleaner.repeatInterval=1800000
# cas.ticket.registry.cleaner.enabled=true

___________________
Juan Quintanilla

Juan Quintanilla

unread,
Mar 28, 2019, 11:42:57 AM3/28/19
to cas-...@apereo.org

Hello,


So after some testing with the DefaultTicketRegistryCleaner on CAS 5.1.9 with JPA ticket registry we have had to disable the DefaultTicketRegistryCleaner because it would take long to remove tickets from the Oracle Database.  We noticed that it would take some time to attempt a cleanup and when it would finally attempt to commit the transactions we would sometimes encounter the error below: 


ERROR [org.apereo.cas.ticket.registry.DefaultTicketRegistryCleaner] - <Unable to find org.apereo.cas.ticket.TicketGrantingTicketImpl with id TGT... 


which would cause the entire transaction to roll back.  We tried having the cleanup process run only on 1 server but we would still see the same issues. This would in turn cause the tickets to accumulate and when the next cleanup would occur we would see even more slowness which would then affect the user logins.  We disabled the DefaultTicketRegistryCleaner and setup some external scripts and have not encountered any issues with ticket cleanup during our load testing.


Has the manner in which DefaultTicketRegistryCleaner is performed changed from version 3.6 to 5.1.9?  When we were running CAS 3.6 we did not have any issues with the DefaultTicketRegistryCleaner.


Thanks!



___________________
Juan Quintanilla


From: Juan Quintanilla
Sent: Tuesday, March 19, 2019 11:51 AM
To: cas-...@apereo.org
Subject: CAS 5.1.9 DefaultTicketRegistryCleaner fails when TGT not found
 

Hello,


We are running 2 CAS 5.1.9 instances with ORacle 12 for the JPA ticket registry and we seeing the following error in the logs:


ERROR [org.apereo.cas.ticket.registry.DefaultTicketRegistryCleaner] - <Unable to find org.apereo.cas.ticket.TicketGrantingTicketImpl with id TGT 


The ticket mentioned in the logs belongs to the 2nd webserver and the cleanup process seems to be only running on the 1st webserver every 30 minutes even though it configured on both.  It never goes to the second webserver. The problem we run into is that because the cleanup process fails to find the ticket everything is rolled back and no ticket is removed.  From what we can find the issue occurs when the TGT is removed from the 2nd server because of a user initiated logout at the same time that the 1st server initiates the cleanup process.  


Has anyone encountered something similar is there maybe a workaround to this so that tickets are cleaned up. Below is the configuration we have for the Ticket Registry and Cleanup Process:


cas.ticket.registry.jpa.ticketLockType=NONE
cas.ticket.registry.jpa.jpaLockingTimeout=3600

#cas.ticket.registry.jpa.healthQuery=select 1
cas.ticket.registry.jpa.isolateInternalQueries=false
cas.ticket.registry.jpa.url=xxxxxxxxx
cas.ticket.registry.jpa.failFast=true
cas.ticket.registry.jpa.dialect=org.hibernate.dialect.Oracle12cDialect
cas.ticket.registry.jpa.leakThreshold=10
cas.ticket.registry.jpa.jpaLockingTgtEnabled=false
cas.ticket.registry.jpa.batchSize=10
#cas.ticket.registry.jpa.defaultCatalog=
#cas.ticket.registry.jpa.defaultSchema=
cas.ticket.registry.jpa.user=xxxxxxxxxx
cas.ticket.registry.jpa.ddlAuto=update
cas.ticket.registry.jpa.password=xxxxxxxxxxx
cas.ticket.registry.jpa.autocommit=false
cas.ticket.registry.jpa.driverClass=oracle.jdbc.OracleDriver
cas.ticket.registry.jpa.idleTimeout=5000
#cas.ticket.registry.jpa.dataSourceName=
cas.ticket.registry.jpa.dataSourceProxy=false
cas.ticket.registry.jpa.pool.suspension=false




Reply all
Reply to author
Forward
0 new messages