APAR status
Closed as program error.
Error description
The problem is that an Enterprise-Application was not able to be
stopped from AdminConsole.
AppServer's javacore shows that the Deadlock situation was
detected.
====================
1LKDEADLOCK Deadlock detected !!!
NULL ---------------------
NULL
2LKDEADLOCKTHR Thread "SoapConnectorThreadPool : 1"
(0x4538C9C0)
3LKDEADLOCKWTR is waiting for:
4LKDEADLOCKMON sys_mon_t:0x45DAA0F8 infl_mon_t: 0x00000000:
4LKDEADLOCKOBJ java.util.HashSet@3065C500/3065C508:
3LKDEADLOCKOWN which is owned by:
2LKDEADLOCKTHR Thread "NotificationService dispatcher : 236"
(0x4969B590)
3LKDEADLOCKWTR which is waiting for:
4LKDEADLOCKMON sys_mon_t:0x45DAA078 infl_mon_t: 0x45FE0228:
4LKDEADLOCKOBJ java.lang.Object@306AA738/306AA740:
3LKDEADLOCKOWN which is owned by:
2LKERROR8 <UNKNOWN THREAD> Thread <unknown thread> (0x22C)
====================
And it shows that 100 of "NotificationService dispatcher"
threads existed. Three of "NotificationService dispatcher"
threads were trying to stop an application, but they and other
most threads were blocked by other Notification thread.
On one hand, it seems that nodeagent were working normally at
that time, from its javacore.
AppServer Trace logged many ConnectorException, but it seems
that it is a victim of the problem. And, SystemErr logged many
Security Exceptions; I am not sure whether they are related to
this problem.
Now, we need to know that why the "NotificationService
dispatcher" generated a deadlock situation, and the way to
solve this problem.
Local fix Problem summary
****************************************************************
* USERS AFFECTED: WebSphere Application Server security *
* users. Stopserver or stop application may *
* hang. *
****************************************************************
* PROBLEM DESCRIPTION: Stop server or stop application may *
* hang *
****************************************************************
* RECOMMENDATION: *
****************************************************************
When shutdown server or stop application, server process may
try to login after Login Module has been destroyed, which in
turn trigger recursive login attempts, and cause system to hang.
When problem occurs, a javacore will show "deadlock" on
NotificationService dispatcher thread, and lots
"NotificationService dispatcher" threads.
Problem conclusion
This is a timing issue where server process login is performed
after login module is destroyed. We add a logic first not to
trigger recursive login, then to have time window for a
validated security subject to survive after login module dies.
The fix for this APAR is currently targeted for inclusion in
fixpack 5.0.2.14. Please refer to the Recommended
Updates page for delivery dates:
http://www-1.ibm.com/support/docview.wss?rs=180&context=SSEQTP
&uid=swg27004980
Temporary fix
test fix provided
Comments
APAR information |
APAR number |
PK08725 |
Reported component name |
WAS NETWRK DEPL |
Reported component ID |
5630A3601 |
Reported release |
00A |
Status |
CLOSED PER |
PE |
NoPE |
HIPER |
NoHIPER |
Special Attention |
NoSpecatt |
Submitted date |
2005-07-12 |
Closed date |
2005-08-30 |
Last modified date |
2005-08-30 |
APAR is sysrouted FROM one or more of the following:
APAR is sysrouted TO one or more of the following:
Modules/Macros
Publications Referenced
|
Fix information |
Fixed component name |
WAS NETWRK DEPL |
Fixed component ID |
5630A3601 |
Applicable component levels |
R00A PSY |
UP |
R00H PSY |
UP |
R00I PSY |
UP |
R00P PSY |
UP |
R00S PSY |
UP |
R00W PSY |
UP |
|