The CLEAN entrypoint is called after a START or STOP fails. It is also called when a CHECK entrypoint returns FAILED for a previously ONLINE resource. The intention is to clean any residual state left behind by the resource/application.
The ABORT entrypoint is not called in scripts, only for C/C++ agents.
Would a hung START/STOP/CHECK also qualify as "failed" or only those that complete (and do not hang) and return a 1 (and not a 0) qualify as "failed"?
Also, if CHECK returns FAILED (or if CHECK is hung), then CLEAN will be called and then START will be called. And
STOP will not be called. Is that right?
- Hung START/STOP/CHECK :- only ABORT is called. In the case of action scripts, a ABORT just kills the corresponding launched script (by scriptagent). CHECK is called after the ABORT.
- If a CHECK after START/STOP or routine CHECK returns FAILED (value 5), then CLEAN is called.
- Yes, if resource was running fine (ONLINE) and then CHECK action returns FAILED, a CLEAN is called first and then the resource is restarted on same node or