Re: monit restart bug/race condition (3.1 and current CVS version behavior)

[Martin Pala]

Christian Hopp wrote:

> On Mon, 10 Feb 2003, Jan-Henrik Haukeland wrote:
>
> I am back after a heavy bronchitis. (-:
I hope you feel better :)

> - In a "restart" situation the resource data "ProcInfo_T" is not
>  correctly reinitialized (zeroed)... e.g. -6324% CPU load
Strange - i tested it today and i wasn't able to replicate it. There was 
this problem in the past (and probably in 3.1 too) but i though it was 
fixed (current CVS version is OK i hope). The restart method is based on 
atomic stop and start methods - stop method does reset of procinfo 
structure:

 /* Reset the proc info object in case of a later restart */
 memset(p->procinfo, 0, sizeof *(p->procinfo));

> For 3.3:
> - I wanna try to implement the process child count and total
>  child process memory.  But that's gonna be tricky... it is necessary
>  to review every process in order to find out the children of a
>  process.  If you want to review a whole "children branch" it even more
>  work (tree structures, ...)
>
> Christian
>
>  
Sounds good :)