dns performance and storeSwapOut again

Dear all,

 I have tracked down my performance problems to be mainly caused by
dns. I have tried pointing to various machines, one of them not
overloaded at all and also setting up a cache-only bind in the same
machine with no results. I am worried about the following traces (snoop
and tcpdump) run on the bind servers I was pointing to:

09:35:42.510000 belinda.rediris.es.domain > proxy.rediris.es.47059:
1710* q: pages.prodigy 1/2/2 . (154) (ttl 64, id 15931)
09:35:42.510000 proxy.rediris.es > belinda.rediris.es: icmp:
proxy.rediris.es udp port 47059 unreachable (DF) (ttl 255, id 43637)

---
proxy.rediris.es -> tais.rediris.es DNS C port=53883
tais.rediris.es -> proxy.rediris.es DNS R port=53833
proxy.rediris.es -> tais.rediris.es ICMP Destination unreachable (Bad
port)
 they repeat very often (one out of ten requests)
 Any hint on what might be the cause of the ICMP unreachable one
milisecond after requesting from that port ??
 Unluckily enough from time to time I keep having the
storeSwapOutFileOpened problem:
(2) No such file or directory
1999/03/04 10:52:18| storeSwapOutFileOpened: Unable to open swapfile:
/cache7®/00/01/0000010D
 this time /cache7 does not even exist neither is referenced anywhere.
ok, I'll tell if I find out anything. Please report anyone having the
same problems.
Best regards,
Javier Puche.