Hi all:

Our DB2 crashes every 1/2 hour, since 4 days ago. The only way to restart it is by opening Control Center or Command Center.
Environment: Db2 V8.1, Windows 2000 Server.
I looked into DB2diag.log and here is part of it

Any Idea?....plase help....urgent!

---------------------------------------------------------------------
2004-10-07-15.42.23.015000 InstanceB2 Node:000
PID:5896(db2syscs.exe) TID:6072 Appid:none
database monitor sqm_object::new Probe:100

DIA8300C A memory heap error has occurred.
ZRC=0x8B0F000B

2004-10-07-15.42.23.234000 InstanceB2 Node:000
PID:5896(db2syscs.exe) TID:5976 Appid:*LOCAL.DB2.02B247194222
base sys utilities sqleagnt_sigsegvh Probe:1 Database:BDGEST

Error in agent servicing application with coor_node:0000 ..

2004-10-07-15.42.23.406000 InstanceB2 Node:000
PID:5896(db2syscs.exe) TID:5976 Appid:*LOCAL.DB2.02B247194222
base sys utilities sqleagnt_sigsegvh Probe:2 Database:BDGEST

Error in agent servicing application with coor_agent_index:7d00 }.

2004-10-07-15.42.23.562000 InstanceB2 Node:000
PID:5896(db2syscs.exe) TID:5976 Appid:*LOCAL.DB2.02B247194222
base sys utilities sqleagnt_sigsegvh Probe:3 Database:BDGEST

Error in agent servicing application with CLIENT PID:3532 3136 5216

2004-10-07-15.42.23.718000 InstanceB2 Node:000
PID:5896(db2syscs.exe) TID:5976 Appid:*LOCAL.DB2.02B247194222
base sys utilities sqleagnt_sigsegvh Probe:4 Database:BDGEST

Error in agent servicing application with INBOUND APPLICATION ID:2a4c 4f43 414c 2e44 4232 2e30 3242 3234 *LOCAL.DB2.02B24
3731 3934 3232 32 7194222

2004-10-07-15.42.23.921000 InstanceB2 Node:000
PID:5896(db2syscs.exe) TID:5976 Appid:*LOCAL.DB2.02B247194222
base sys utilities sqleagnt_sigsegvh Probe:5 Database:BDGEST

Error in agent servicing application with INBOUND SEQUENCE NUMBER:0100 ..

2004-10-07-15.42.24.125000 InstanceB2 Node:000
PID:5896(db2syscs.exe) TID:5976 Appid:*LOCAL.DB2.02B247194222
base sys utilities sqleagnt_sigsegvh Probe:6 Database:BDGEST

is extshm parameter is set to 'on' for this database? what's the bufferpool size? you may need to increase dbheap..need info. I had this problem before

also, after db2stop check if there are any processes hanging and holding the memory..
do ipcs and if u see any memory held under the instance name clear it with ipcrm

Hi Chavadb2:

Thanks for your replay. My comments:

1) "extshm": How do I ckeck it ?
2) "increase dbheap": there are a few heaps..which one should I increase?
3) "ipcs and if u see any memory held under the instance name clear it with ipcrm": How do use IPCS and IPCRM. Are these from "Windows Task Manager"?

4) Update:
I have increased MON_HEAP_SZ from 60 to 85. The problem happens less
often....Is this the right way?

Thanks...

sorry these are unix commands.
if extshm is on it will be added to your .profile and db2profile
this is used to allocate more shared memory segments to java apps
yes.. increase the mon_heap if you turned on any of the monitoring switches

dbheap should be set greater than what you get from the following formula

For Version 7

In bytes:
(logbufsz * 4K)
+ (catalogcache_sz * 4K)
+ 8K per table space
+ 4K per table
+ 100 bytes per page of Extended Storage

In pages:
logbufsz
+ catalogcache_sz
+ 2 per table space
+ 1 per table
+ 1 per 40 pages of Extended Storage

If buffer pool monitoring is enabled, you will need to add the following:

216 bytes
* (number of database agents + number of prefetchers + number of page cleaners)
* (maximum table space ID)

For FixPaks 4 through 7 inclusive, use 416 instead of 216 in the above formula.

Note that the buffer pool monitoring requirement is only a concern when there is a large number (more than 100) of table spaces.

For Version 8

In bytes:
(logbufsz * 4K)
+ 10K per table space
+ 4K per table
+ ( 1K + 4*extents used ) per range clustered table (RCT)


Description of components of dbheap :

logbufsz: Memory that is allocated in its entirety on database activation.
catalog cache (Version 7 only): This is allocated in its entirety on database activation. The catalog cache loads descriptions of each table from the catalog as they are accessed. The memory that is actually used never increases or decreases. Table descriptions are only removed when there is no room left in the cache and a table whose description is not already loaded into the cache needs to be accessed.
table space control information: Control information for all table spaces is loaded on database activation, or as table spaces are created; 8K per table space in Version 7, 10K in Version 8.
object control information: Applies to tables, indexes, or LOBs, approximately 4K per table, depending on the table and index definitions. Object control information is loaded as tables are accessed, so the allowance for tables only needs to include accessed tables. Once a table is accessed, this information stays in memory until the database is recycled.
memory for buffer pool monitoring (Version 7 only): This is volatile memory. The maximum is 216 bytes * (agents + prefetchers + pagecleaners) * maximum_tablespace_id. The 216-byte value was increased to 416 bytes in FixPak 4 for internal performance monitoring purposes, and reverted to 216 bytes in FixPak 8 (APAR IY32333).
range clustered table (Version 8 only): RCTs are supported as of Version 8, FixPak 4. The memory requirement for an RCT is in addition to the object control information.

If you have fixpak 4a or earlier, this sounds like APAR IY50445. Fixed in FP5, but FP 6a is your best bet.

Ok. Thanks Marcus and Chavadb2.
Then, I have a lot of thinks to check .