How to collect GDB stack trace from a SpectroSERVER core dump on Linux
Article ID: 191615
Updated On:
Products
Issue/Introduction
When a Spectrum process crashes on Linux, we may ask you to process the core file on the system where is was generated.
This is because the version and files referenced only exist on your system.
The tool you would use is GDB.
Environment
DX NetOps Spectrum: All Supported Versions
Resolution
Please install the Linux GNU GDB Debugger on the system, if not already installed, by running the following command as the root user:
yum install gdb
The SpectroSERVER core file will be in the $SPECROOT/SS directory.
Example of a core file generated by the SpectroSERVER process:
core.<pid> for example core.3621
core.<details> for example core-SpectroSERVER-sig11-userXXXX-groupXXXX-pid2079677-time1705969295
These steps will provide the "where", "info threads" and "thread apply all bt full" information from the GDB output.
1 - Login, su or sudo to the root user
2. cd to the $SPECROOT/SS directory and run this using the core file name
3. Run gdb:
gdb -ex "where" -ex "info threads" -ex "thread apply all bt full" -ex "quit" SpectroSERVER CoreFileName > /tmp/gdb_where.out 2>&1
Be sure to replace the "CoreFileName" with the actual name of the core file.
For example:
gdb -ex "where" -ex "info threads" -ex "thread apply all bt full" -ex "quit" SpectroSERVER core.3621 > /tmp/gdb_where.out 2>&1
Upload the /tmp/gdb_where.out file to your Broadcom case.
Optionally you can run the commands one at a time where <CORE> is the filename of the core file:
echo `date` >> gdb.output ; echo "where output" >> gdb.output;
gdb -ex "where" -ex "quit" SpectroSERVER <CORE> >> gdb.output;
echo "info threads output" >> gdb.output;
gdb -ex "info threads" -ex "quit" SpectroSERVER <CORE> >> gdb.output;
echo "thread apply all bt full output" >> gdb.output;
gdb -ex "thread apply all bt full" -ex "quit" SpectroSERVER <CORE> >> gdb.output;
3 - Upload the gdb.output file to the case for analysis.
NOTE: any process can generate a core file on Linux, in this example we are using the SpectroSERVER,
but the same steps apply to any core, just run it from the same directory as the core and use the process name that crashed.
Additional Information
To run the above steps manually:
a) As root, cd to the $SPECROOT/SS directory
b) If the file in the SS directory has not been renamed, rename the core
file to something other than core so what if GDB crashes it will not
overwrite the core file (ie. mv core core.1)
c) Execute the following and a few lines of data will print to the screen:
d) gdb SpectroSERVER <core_file>
Example: gdb SpectroSERVER core.1000
Type the following to redirect the output to a text file (gdb.txt)
(gdb) set logging on
You can tell it which file to use:
(gdb) set logging file my_gdb_object.log
And you can examine the current logging configuration:
(gdb) show logging
e) Type "where" (without the quotes) and hit enter
f) Several lines of data (the contents of the stack) will print out
g) Type "info threads" (without the quotes) and hit enter
h) Several lines of data (the contents of the stack) will print out
i) Type "thread apply all bt full" (without the quotes) and hit enter
j) Type $Q to quit out of gdb
k) Upload the $SPECROOT/SS/gdb.txt file.
Feedback
