Ross J. Reedstrom wrote:
On Mon, Jun 15, 2009 at 12:16:27PM -0800, Chris Robertson wrote:
Using "squidclient mgr:server_list" there is a "AVG RTT" value. In
mgr:digest_stats there is "icp.query_median_svc_time".
Which are both useful values. However, they don't tell me what value the
server is using for the dynamic timeout. AFAICT, it's being kept on a
per-server basis. The mystery is that I've got a load balancing setup,
How did you set load balancing up?
and squid seems to be favoring servers that server_list claim an AVG RTT
in the 200-500 ms range, when other, less favored servers are showing
5-20 ms. Walking the code now (which is much easier once I realized my
editor's tabstop was set for the python-default 4 spaces, not c-friendly
8. Oops) Anyone got pointers to any sort of higher-level design for any
of this? I've spent significant time poking around the wiki and FAQ, and
haven't surfaced much.
Ross
Chris
