How to set a large trace buffer in DB2 LUW

Body

This might be commonly used now a days.  But,  questions on this come up over and over still.   So,  trying to put  few lines on that.

Generally,  the  db2trc  has a  maximum  buffer size  of  128m  and it can differ  based on  platforms/levels  too. 

In certain platform and levels  it cannot set to even 128m unless it's  set  before a db2start.

Also,   if  a trace buffer is set to a smaller value  example 64m  for the first  time  after  a  db2start  it  cannot be increased or changed to a different  value  unless the db2 is recycled.

There are good  blogs and technotes  where  the reason is explained in details.

So,  here   we  want to know how a trace buffer could be changed to  a  higher value than  128m

To  do that,  first  a  db2 registry parameter  has to be set  which  is called,  DB2TRC_DEF_BUFFSIZE

Example,

To  set a  512 MB of trace buffer,

db2set DB2TRC_DEF_BUFFSIZE=512m

To  set a  1 GB of trace buffer,

db2set DB2TRC_DEF_BUFFSIZE=1g

Then the db2 needs to be restarted to make that  effective,

db2stop/db2start

After that  the trace could be set  using (example 1 GB),

$ db2trc on -l 1G
Trace is turned on

$ db2trc info
Marker                  :  @TRACE@
Trace version           :      7.0
Platform                : AIX 64BIT
Build level             : s141128
maxBufferSize           : 1073741824 bytes (1024 MB)
auxBufferSize           : 0 bytes (0 MB)
allocationCount         : 2
DB2TRCD pid             : 0
Trace destination       : <shared memory buffer>
numSuspended            : 0
Trace starting time     : 2017-02-06-12.59.59.394927-300

Buffer size             : 1073741824 bytes (1024 MB)
Allow buffer to wrap    : yes
Mask                    : *.*.*.*.*
Timestamps              : disabled
PID.TID mask            : all
Fixed data mask #1      : all
Fixed data mask #2      : all
Max system errors       : infinite
Treat this rc as sys err: none
Member mask             : none
Application handle mask : none
Application ID mask     : none

Point to be noted,    even after setting this to 1GB using the registry if   for the very first time after a  db2start  the buffer size is set to a smaller value it cannot be  changed  back unless the db2 is recycled again.

Sometimes   question arise  about why  not use  -f   option  just to  put  unlimited  trace  in a file.   The answer to that  usually will be,    with  the file option  there could be  additional overheads for the  output  file I/O  which  is not suggested to be used  while  a  performance  issue is being  diagnosed. If it's not related to performance  troubleshooting then the file option is o.k. to be used.  But,  -f option is still to be used with  caution as that can create huge output file resulting disk  full   or sometimes that much large file don't become useful for analysis.