When a database is activated, if there is insufficient memory to
start all the defined user buffer pools, the database will still
be activated with tiny system buffer pools only (also commonly
referred to as 'hidden' buffer pools) in order to allow for an
administrator to rectify the situation. In such a scenario, the
db2diag.log will contain messages about the insufficient memory
condition (as seen further below), and the ADM6073W warning
message will be logged into both the db2diag.log and admin logs.
When this same condition occurs during a START HADR AS STANDBY
operation or ACTIVATE DATABASE command on a HADR standby
database, the standby database is incorrectly allowed to start
with only the system buffer pools. The command should have been
rejected with error SQL1768 reason code 13. It is not desired
behavior to allow the use of HADR standby database with only
system buffer pools because it would drastically impact standby
log replay performance.
Sample messages in db2diag.log:
FUNCTION: DB2 UDB, SQO Memory Management, sqlogmshr, probe:149
MESSAGE : ZRC=0x850F00C5=-2062614331=SQLO_NO_INS_MEM
"Insufficient instance_memory to allocate new memory"
DATA #2 : String, 56 bytes
Insufficient INSTANCE_MEMORY to allocate new memory set.
...
FUNCTION: DB2 UDB, base sys utilities,
sqeLocalDatabase::FirstConnect, probe:10000
MESSAGE : ZRC=0x850F00C5=-2062614331=SQLO_NO_INS_MEM
"Insufficient instance_memory to allocate new memory"
DATA #1 : String, 283 bytes
Failed to allocate the database shared memory set.
The configured DATABASE_MEMORY plus smaller overflow may have
exceeded INSTANCE_MEMORY or the maximum shared memory on the
system.
Attempting to start up with only the system buffer pools.
Database shared memory set size is (bytes):
...
FUNCTION: DB2 UDB, buffer pool services, sqlbAssignBufferPool,
probe:2
MESSAGE : ADM6073W The table space "SYSCATSPACE" (ID "0") is
configured to use
buffer pool ID "1", but this buffer pool is not active
at this time.
In the interim the table space will use buffer pool ID
"4099". The
inactive buffer pool should become available at next
database startup
provided that the required memory is available.
...
With this APAR fix, behavior changes as follows:
If the HADR Standby database encounters the condition
described above, where there is insufficient memory to allocate
all buffer pools, it will instead fail with SQL1768N reason code
13 error.
"SQL1768N Unable to start HADR. Reason code = 13.
13 - An HADR standby database cannot start without at
least one regular buffer pool. A buffer pool could not
be started due to insufficient memory."
On regular db or a HADR primary, a warning message SQL1478W
will continue to be returned with db activated successfully.
"SQL1478W The defined buffer pools could not be started.
Instead, one small buffer pool for each page size
supported by the database manager has been started." |
