Snap data collection
You can collect the IBM Spectrum® Scale snap data from the Ambari GUI. The command is run by the IBM Spectrum Scale Master, and the snap data is saved to /var/log/ambari.gpfs.snap.<timestamp> on the IBM Spectrum Scale Master node.
By default, the IBM Spectrum Scale Master runs the following command:
/usr/lpp/mmfs/bin/gpfs.snap -d /var/log/ambari.gpfs.snap.<timestamp> -N <all nodes> --check-space --timeout 600
Where <all nodes> is the list of nodes in the IBM Spectrum Scale cluster and in the Ambari cluster. The external nodes in a shared cluster, such as ESS servers, are not included.
Note:
- If your cluster has IBM Spectrum Scale™ file system version 4.2.2.0 and later, gpfs.snap will include the --hadoop option.
- If you run the Collect Snap Data through Ambari GUI, the Ambari logs will be captured into a tar package under the /var/log directory. The base gpfs.snap --hadoop option command does not capture the Ambari logs. The Ambari logs are only captured by clicking, in the Ambari GUI.
You can also override the default behavior of this snap command by providing the arguments to the
gpfs.snap command in the file
/var/lib/ambari-server/resources/gpfs.snap.args. This works only if you are
running the gpfs.snap on the command line. For example, if you wanted to write
the snap data to a different location, collect the snap data from all nodes in the cluster, and
increase the timeout. You can provide a gpfs.snap.args file option similar to
that in the following
example:
# cat /var/lib/ambari-server/resources/gpfs.snap.args
-d /root/gpfs.snap.out -a --timeout 1200The Ambari snap data tar package is stored as /var/log/ambari.mpack.snap.<TIMESTAMP>.tar.gz on the IBM Spectrum Scale Master node.
The Ambari snap captures the following information:
- From all Ambari client:
- /var/log/hadoop/root/* - /var/lib/ambari-agent/data/ - /var/log/ambari-agent/ambari-agent.log - From Ambari-server:
- /var/log/ambari-server/ambari-server.log - /var/run/ambari-server/stack-recommendations/
