httpd-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Dean Gaudet <dgau...@arctic.org>
Subject dynamic timeout management
Date Fri, 25 Apr 1997 21:47:24 GMT
Dunno how many of you ran into this today, I'm guessing a few of you did. 
Some goon NSP re-advertised (and black holed) about 25000 extra routes
today (normal defaultless routing table size is 45000).  This caused some
extreme havoc on my servers @hotwired.  Connections were made, maxing out
the MaxClients, and they were *all* stuck in Write state, making the
server stop responding.

I've had this happen before... and I'm beginning to curse asymmetric
routing (hotwired has 3 providers).  Maybe it's not as much a problem
without the asymmetry.

Anyhow I was toying with the idea of dynamically changing the timeouts
such that when this is going on the server is more aggressive at ditching
connections.  But since all my monitoring gear is at least T1 connected
I'd probably see things as "ok" from my monitors but users wouldn't be
able to get anything.

Any input? 

Dean


Mime
View raw message