Hello all!
Apologies for the lenght of the message...
I'm running a cache hierarchy of 22 linux boxes with Squid
1.1.21 installed (upgrading is not very simple since squids
have different configurations and cpus are spread in
different locations in Italy, from Milan to Sicily).
Squid cpus are 2xPentium II, 300 MHz CPU with 512 MB RAM and
34 GB disk (5 to 8 reserved to the squid cache), leaf
proxies are 1xPII 350 MHz, 512MB RAM, 8GB SCSI-II disk (5 GB
reserved for squid-cache).
The entire cluster serves 4000+ users with peaks of 500K
requests per day on main proxyes approx. 250Krequests leaf
caches (1 to 4 GB transferred per day on main proxies).
The network mesh provides connectivity thru 16 frame-relay
(64Kb) network connections and main proxies are connected on
an internal 100 MB LAN.
Internet connection is a single 256 Kb network that main
proxies reach thru a firewall: main proxies are a mesh of
four, connected each other via a switching hub and to the
firewall via a socksified (SOCKS 4) connection.
The squid mesh operates very well under heavy load
conditions, except in case of poor connection to the
Internet: when the internet connection gets slow, main
proxies appear to run in a race condition (CPU load gets at
99.9 % and squid stops accepting requests until internet
connection remain poor.
Logs and debugging seems to be ok: anyone has hints or
suggestions ?
Thank you all for your help and greetings from Italy!
Alessandro
-- Alessandro "Alex" Bruciamonti alessandro@sys-net.itReceived on Wed Mar 31 1999 - 06:47:51 MST
This archive was generated by hypermail pre-2.1.9 : Tue Dec 09 2003 - 16:45:37 MST