Whilst trying to model mozilla downloads, with fast LAN clients/servers, the
memory used by
Lusca-HEAD used grew unbounded when the disks fell behind in handling swapout
traffic.
I'm guessing that fwdCheckDeferRead() is running the server traffic at full
speed because the client-
side is able to sink the traffic quickly.
The change should probably be a twiddle to only allow the swapout offset to
fall a certain distance
behind the slowest client, rather than the default behaviour which is "however
fast it is".
I wonder if this is the same behaviour in Squid-2.HEAD, Squid-2.6/Squid-2.7 and
Squid-3.x ..
Original issue reported on code.google.com by adrian.c...@gmail.com on 8 Apr 2009 at 12:15
Original issue reported on code.google.com by
adrian.c...@gmail.com
on 8 Apr 2009 at 12:15