Re: [Condor-devel] cost of DC timers we don't need?

[Todd Tannenbaum <tannenba@xxxxxxxxxxx>]

Derek Wright wrote:
> part of my starter hacking involves moving the periodic update logic 
> higher up in the JIC class hierarchy, since any starter might define 
> HOOK_UPDATE_JOB_INFO.  currently, the code is a little dumb, in that it 
> always registers the timer, and the timer handler itself invokes a 
> method to do the update (which decides if there's a hook to invoke or 
> not).  this is because the JICShadow *always* wants the timer (for the 
> shadow RSC), but it's the base class that's now managing the timer (so 
> it can be shared).
>
> do we care about a DC timer going off that doesn't do anything in the 
> case of non-shadow starters that don't define this hook?
>
> i could add another JIC method like "needsPeriodicUpdate()", which 
> JICShadow always returns true, and the  others just use the base class's 
> version which checks to see if there's a periodic hook defined.
>
> is it worth the trouble?
>
> thanks,
> -d


probably not, unless the timer is going off real often (which i guess is 
not the case above).

the only cost of a timer firing in an otherwise idle dc process is : (a) 
crap written into the starterlog depending upon logging level, and (b) 
it may prevent the starter from being more fully swapped out by the OS. 
 no big deal imho.



--
Todd Tannenbaum                       University of Wisconsin-Madison
Condor Project Research               Department of Computer Sciences
tannenba@xxxxxxxxxxx                  1210 W. Dayton St. Rm #4257
Phone: (608) 263-7132                 Madison, WI 53706-1685