sparse tail fails when logs are missing (make it fail more gracefully?)

[tdhopper]

sparse tail fails if any of the bash patterns (worker* supervisor* access* metrics* topo_name_*) is empty. This is probably an indication that something bad has happened, so maybe failure is the correct behavior.

Even if failure is the correct behavior, perhaps we could improve the error message from something more clear than:

 Fatal error: run() received nonzero return code 2 while executing!

Requested: cd /opt/storm/current/logs && ls worker* supervisor* access* metrics* topo_name_*
Executed: /bin/bash -l -c "cd /opt/storm/current/logs && ls worker* supervisor* access* metrics* topo_name_*"

For example, streamparse might tell which log files are missing?

[amontalenti]

Maybe we should switch this to use find rather than shell globbing to avoid the error. I don't think this should throw an exception at all.

Also, for very big log file directories, find is probably more efficient than globbing.

[kbourgoin]

It definitely shouldn't throw an exception. Consider a larger storm cluster of 20 or so nodes. We wouldn't have every topology on every machine, so that tail command needs to work even if the topology isn't there.