start_instance(s) and stop_instance(s) add provision job (as run_command and handle_git_command)
scale_instance and create_release set retired on instances.
create_instance insert a new instance.
I think there is a confusion between two different concepts:
- resistant erosion processes
- dynos
Resistant erosion processes (REP) are setup by the user with the ps command.
ps:scale web=2, --> create 2 REPps:scale web=0 -> destroy every REP of type webps:stop --> destroy every REPps:restart --> do not touch the REP, and restart the dyno- set idle --> do not touch the REP, and stop the dyno
- http request but REP is idle --> do not touch the REP, and start the dyno
- midnight restart -> do not touch the REP, and restart the dyno
- application crash --> store crash info in the REP and start a new dyno if number of crash is not too much.
- dynohost crash --> do not touch the REP, and start new dynos
The role of an REP is to make sure a dyno is started even if crash happens.
May be useful for #7
start_instance(s)andstop_instance(s)add provision job (asrun_commandandhandle_git_command)scale_instanceandcreate_releaseset retired on instances.create_instanceinsert a new instance.I think there is a confusion between two different concepts:
Resistant erosion processes (REP) are setup by the user with the ps command.
ps:scale web=2, --> create 2 REPps:scale web=0-> destroy every REP of type webps:stop--> destroy every REPps:restart--> do not touch the REP, and restart the dynoThe role of an REP is to make sure a dyno is started even if crash happens.
May be useful for #7