httpd-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Chuck Murcko <>
Subject Re: mod_proxy changes
Date Mon, 27 May 1996 19:41:29 GMT
Robert S. Thau liltingly intones:
> One more note... a consequence of this item, which was already on
> Chuck's list...
> 10) Make it garbage collect in the background, not while someone is
>         waiting for a response! 
> is that not only does the server process make the client wait for the
> collection to complete, but its data segment grows large enough to hold
> the entire map of the cache (which files, how old), and then stays that
> large for as long as the server process itself keeps going.  (This is
> similar in spirit to the problems Rasmus was having with temps piling
> up from long PHP runs).
> Forking, detaching, and doing a collection in the child (if one isn't
> already in progress) would take care of both problems; it might be
> worthwhile.
This was my thought too, as soon as I fired up the proxy and looked at
the allocated memory size. For now, I'm just setting a finite child
life so it doesn't get out of hand. It actually might be cleaner to
fire up a GC daemon when the proxy starts up rather than fork/exec for
each child, but that's a stage 2 modification.

> (Also, if putting a subdirectory structure into the cache arena wasn't
> on the TODO list, you might consider that as well --- the current
> structure is likely to get quite cumbersome in any industrial-scale
> use...).
I believe this is already done by Ben.

Chuck Murcko	N2K Inc.	Wayne PA
And now, on a lighter note:
Polymer physicists are into chains.

View raw message