If you already contacted support,
continue to the component-specific MustGather information. Otherwise,
click: MustGather: Read First for all WebSphere Application Server
Products.
Use this information to trace systems management,
connection manager, object request broker (ORB), and/or security
components that run as subcomponents and are part of the administrative
server. This technique is especially helpful when multiple days of
execution are required to recreate a problem. In such cases, the enabling
of tracing using normal files within the admin.config file is
inappropriate because of the volume of trace data that can be
generated.
With this tracing technique, the trace files that
are generated are cyclically reused, so the portion of the trace
associated with a particular situation can be captured. This allows the
last N trace records to be saved. You will need the attachments at the
bottom of this page.
This has been tested on AIX. It should work on other
Unix platforms as well, but has not been tested on other
platforms.
- Create a named (FIFO) pipe somewhere in the file system. For example:
mkfifo /usr/WebSphere/AppServer/wsFIFO
If you use ls -lgo to examine this, you will see
something like:
-> ls -lgo
/usr/WebSphere/AppServer/wsFIFO
prw-r--r-- 1 0 Jul 26 12:11 /usr/WebSphere/AppServer/wsFIFO
- Edit admin.config and set the administrative server trace output to
point to the FIFO:
com.ibm.ejs.sm.adminServer.traceOutput=/usr/WebSphere/AppServer/wsFIFO
- Set the administrative server trace string. This is done in
admin.config file by adding the following line:
com.ibm.ejs.sm.adminServer.traceString=com.ibm.ejs.*=all=enabled
In the above, ejs tracing is an example; you can add other trace strings
for connection manager, security or ORB that are pertinent to your
situation. You can use multiple trace strings separated by a colon ( : ).
- Stop all WebSphere-related processes.
- Open a terminal session, change to the directory where you copied the
fifoReader program (attached), then start the
fifoReader program as in:
fifoReader base_namepath_to_FIFOnbr_output_filesmax_lines
Where:
base_name is used to create the output file names. For
example, if you specified "/tmp/ajay", then the output files are named
/tmp/ajay01, /tmp/ajay02, and so forth.
path_to_FIFO tells where (in the file system) the FIFO is
located. This must correspond to what you entered in step 1.
nbr_output_files is the maximum number of output files to
be created. When this limit is reached, the first file is reused.
max_lines is the maximum number of lines to write in each
file before moving on to the next file
For example:
fifoReader
/usr/WebSphere/AppServer/logs/fifoTrace /usr/WebSphere/AppServer/wsFIFO 8
100000
This creates files like:
/usr/WebSphere/AppServer/logs/fifoTrace01
/usr/WebSphere/AppServer/logs/fifoTrace02
/usr/WebSphere/AppServer/logs/fifoTrace03
/usr/WebSphere/AppServer/logs/fifoTrace04
/usr/WebSphere/AppServer/logs/fifoTrace05
/usr/WebSphere/AppServer/logs/fifoTrace06
/usr/WebSphere/AppServer/logs/fifoTrace07
/usr/WebSphere/AppServer/logs/fifoTrace08
and each file has no more than 100,000 lines. Note that you can
choose to either raise the number of files or the number of lines;
however, it is strongly recommended that you use no fewer than 4 files of
100,000 records each.
Start the WebSphere Administrative Server, and watch for heap dumps. At
this point, save the output files so that they are not overwritten. |