Hello,
DB2 V8 LUW FP12 + Linux.
I have just turned on DB2's connection concentrator feature.
MAX_COORDAGENTS=1100 and MAX_CONNECTIONS=1500.
Two hours after restarting the instance, the db stopped responding.
Unfortunately, there was absolutely no message in db2diag.log
(diaglevel 3).
I was even unable to simply locally connect to the DB (no network
issues). A simple "list applications show detail" never returned any
results.
To bring the system back online, I had to kill -9 all db2 processes and
restart the instance.
I am puzzled. Why has DB2 stopped responding? Now everybody wants to
turn off connection concentrator.
I am actually interested in finding out how to troubleshoot a hung
system when you are unable to connect to the DB, and see absolutely
nothing in db2diag. Can anyone suggest any tools?
Thanks in advance.
Larry - 29 Jun 2006 19:37 GMT
> Hello,
>
[quoted text clipped - 22 lines]
>
> Thanks in advance.
First of all, what makes you think that this was definitely related to
the connection concentrator? It might be, but it might not be.
Secondly, why wouldn't you stress test this prior to putting it into
production?
Lastly, your best bet would probably be to open up a PMR with IBM
support as soon as this happens again ... before you do anything to
attempt to restore the system.
FYI. Hangs can occur anyway ... in the application or in the OS kernel
... and can be one of the most difficult sitations to diagnose. The
service team usally needs a callstack if I'm not mistaken. You may not
get that by using kill -9.
Larry Edelstein
Ian - 29 Jun 2006 21:48 GMT
> I am actually interested in finding out how to troubleshoot a hung
> system when you are unable to connect to the DB, and see absolutely
> nothing in db2diag. Can anyone suggest any tools?
The db2pd command is very helpful in situations like this, provided that
you can understand what you're looking at.
As Larry suggested, you best option is to open a PMR with IBM.
Thanks -
Ian
Michel Esber - 30 Jun 2006 14:06 GMT
Thanks guys,
The problem has happened again and I instructed my team to open a PMR.
-Michel
> > I am actually interested in finding out how to troubleshoot a hung
> > system when you are unable to connect to the DB, and see absolutely
[quoted text clipped - 8 lines]
>
> Ian