some host "marked bad"

I'm getting these messages in cache.log very often (squid-1.NOVM.22 in
SCO Unixware 2.1.3):

the connection to the parents gets lost, after a while my cache recovers
the connection, then it looses the connection again, all this
happenning every 10 or 15 minutes:

1998/09/09 16:58:31| TCP connection to bo.cache.nlanr.net/3128 failed
1998/09/09 16:58:45| uc.cache.nlanr.net(141.142.121.5) marked bad
....
1998/09/09 16:58:45| TCP connection to uc.cache.nlanr.net/3128 failed
1998/09/09 16:58:47| bo.cache.nlanr.net(192.52.106.30) marked good
1998/09/09 16:58:47| bo.cache.nlanr.net(192.52.106.31) marked good
1998/09/09 16:58:53| uc.cache.nlanr.net(141.142.121.5) marked good
1998/09/09 16:59:07| TCP connection to bo.cache.nlanr.net/3128 succeeded

1998/09/09 16:59:15| TCP connection to uc.cache.nlanr.net/3128 succeeded

.........
1998/09/09 17:15:22| bo.cache.nlanr.net(192.52.106.30) marked bad
1998/09/09 17:15:22| bo.cache.nlanr.net(192.52.106.31) marked bad
.....
1998/09/09 17:15:24| bo.cache.nlanr.net(192.52.106.30) marked good
1998/09/09 17:15:30| bo.cache.nlanr.net(192.52.106.31) marked good
.....
1998/09/09 17:15:58| TCP connection to bo.cache.nlanr.net/3128 succeeded

I'd like to know if there's something to tune, I've three parents (two
from NLANR, and a third-one in Argentina, which is a squid-1.1.14
without ICP):

cache_host uc.cache.nlanr.net parent 3128 3130
cache_host bo.cache.nlanr.net parent 3128 3130
cache_host 200.10.202.4 parent 1280 7

TIA

carlos horowicz
argentine foreign office
carlos@atina.ar
Received on Wed Sep 09 1998 - 11:45:48 MDT