lock monitoring

lock monitoring

Post by joe bu » Fri, 15 Nov 2002 03:34:44



Hello,

Solaris 2.8
ASE 12.5
Meta Frame server farm (6)

We have about 125 concurrent PowerBuilder users funneled through a
Metaframe server farm.   The clients login to MF server and launch an
instance of PB.  Sometimes an ungracefull exit from the client will
leave a instance of PB* on the MF server  still logged into ASE
(MF zombies).  Last night we had a couple MF zombies* around
with shared locks on objects.  We identified them by noticing that
they were sitting on objects with shared locks and not moving off.

Does anyone have a suggestion for identifing such beast?  I thought
about looking at CPU but these guys were not consuming cpu cycles.
The only thing I could think of was looking at the locked pages and
see if a spid was sitting on a certain page for more than a couple
minutes or some interval.  Any suggestions would be appreciated.
JAB

 
 
 

lock monitoring

Post by Anthony Mandi » Sun, 17 Nov 2002 11:12:33



> We have about 125 concurrent PowerBuilder users funneled through a
> Metaframe server farm.   The clients login to MF server and launch an
> instance of PB.  Sometimes an ungracefull exit from the client will
> leave a instance of PB* on the MF server  still logged into ASE
> (MF zombies).  Last night we had a couple MF zombies* around
> with shared locks on objects.  We identified them by noticing that
> they were sitting on objects with shared locks and not moving off.

> Does anyone have a suggestion for identifing such beast?  I thought
> about looking at CPU but these guys were not consuming cpu cycles.
> The only thing I could think of was looking at the locked pages and
> see if a spid was sitting on a certain page for more than a couple
> minutes or some interval.  Any suggestions would be appreciated.

        Probably comparing sysprocesses entries to see if there's been
        any activity. The real issue, of course, is why doesn't the
        Citrix rubbish clean up by shutting down processes when a
        client exits?

-am     ? 2002

 
 
 

lock monitoring

Post by joe bu » Wed, 20 Nov 2002 01:45:58


The real issue, of course, is why doesn't the

Quote:>    Citrix rubbish clean up by shutting down processes when a
>    client exits?

> -am        ? 2002

Thats true but I don't have much control over that.  It's operated by
a different department on the other side of the states.  So I am left
to defend for myself.

jab

 
 
 

1. Lock monitor thread polling period

Hello,

The MSSQL docs state that for deadlock detection, the lock monitor thread
will periodically wake up and check for locks. If a process is still waiting
on the same lock upon the second check, a deadlock check will be run.Does
anyone know how long the lock monitor sleeps between checks and/or how I can
find this out?

Thanks,
Dave

2. Oracle 7.3.2 NT Listener

3. Lock Monitor: Deadlock Chain

4. xp_send mail and the @query

5. system processes, e.g., LOCK MONITOR, LAZY WRITER, LOG WRITER, CHECKPOINT SLEEP

6. Slow DROPing Table

7. Table/Page Lock Monitor

8. Updatable recordset from parameter stored procedure

9. Sqlserver lock monitor crash dump

10. Database lock monitoring tools

11. Lock Monitor Question

12. Lock Monitoring question

13. Any decent utilities for lock monitoring?