Re: monit monit_http.c

[Jan-Henrik Haukeland]

Rory Toma <address@hidden> writes:

> Cool, that would've taken me awhile to find. (not being a thread
> expert...)

As I said, I'm not 100% confident that this will actually fix the
problem.

The root cause (I think) for the existence of a running rogue http
thread may be that the monit daemon crashed and that the http thread
is hanging on the accept() call (normally, on shutdown the monit
daemon close the server socket so the http thread cannot hang on the
accept call and exit.) In theory the pthread system should kill the
http thread when the main thread dies but if the monit daemon thread
simply crashes it may cause a hiccup in the pthread system. (At least
on Linux where the quality of the pthread implementation is not
perfect). Anyway, putting the http thread in a detached state may help.

-- 
Jan-Henrik Haukeland