Timeouts for get_next_session not reported correctly

[tisba]

Although ts_launcher:do_launch/1 should catch this properly (in

tsung/src/tsung/ts_launcher.erl

Line 395 in bb804e8

[{'EXIT', {timeout, _ }}] ->

), I still saw this recently:

=INFO REPORT==== 21-May-2015::09:08:53 ===
         ts_launcher:(3:<0.177.0>) get_next_session failed for unexpected reason [{'EXIT',
                                                                                   {timeout,
                                                                                    {gen_server,
                                                                                     call,
                                                                                     [{global,
                                                                                       ts_config_server},
                                                                                      {get_next_session,
                                                                                       "ip-172-31-13-176",
                                                                                       1}]}}}], abort !

I'd like to provide a patch, but I do not understand why this can happen ([{'EXIT', {timeout, _ }}] seem to be correct to match the error reported above. I'm also not sure how to reproduce this reliably.

/cc @nniclausse

[tisba]

I also saw this error, where I'm actually not sure what it actually means:

=INFO REPORT==== 21-May-2015::13:37:20 ===
         ts_launcher:(3:<0.484.0>) get_next_session failed for unexpected reason [{'EXIT',
                                                                                   {noproc,
                                                                                    {gen_server,
                                                                                     call,
                                                                                     [{global,
                                                                                       ts_config_server},
                                                                                      {get_next_session,
                                                                                       "ip-172-31-0-122",
                                                                                       1}]}}}], abort !

[tisba]

Wow, nice! Could you please explain what the other error means, @nniclausse?

[nniclausse]

i think it means that the ts_config_server does not exist from the ts_launcher point of view. ts_config_server is a globally registered process and tsung forces a global:sync before starting, so it should not happend in theory :) Now i don't know how the global registration stuff works when either the controller or one of the client nodes gets overloaded

[tisba]

Thanks. The cluster was quite overloaded, so I guess this is could be the reason. I was just looking through the logs to look for issues and thought I report anything I find.