I am getting an odd delay on traceroutes on a private connection
through a private vlan switch , being used for oracle 10G
interconnect.

The traceroutes work, but there is a 5-10 second delay in the
response.
Traceroute -v shows that it is trying the loopback first , failing
multiple times, then returning the good response. Pings are normal
duration, entstats don't show any errors. All the other interfaces on
the cluster behave normally.

Both ip's are in the hosts tables on each box, it doesn't look like a
name resolution issue..behaves the same way with numeric or named
address.

Anyone know what would cause it to hit and fail on 127.0.0.1 first? (
note that a direct traceroute to 127.0.0.1 is ok)

sample output plus partial netstat -r follows. Thanks.


[ROOT:apadb255n1]/# traceroute -v apadb255n2-private
trying to get source for apadb255n2-private
source should be 192.168.248.144
traceroute to apadb255n2-private (192.168.248.145) from 192.168.248.144
(192.168.248.144), 30 hops max
EMSGSIZE 32768
EMSGSIZE 32748
EMSGSIZE 17914
EMSGSIZE 17894
EMSGSIZE 16384
EMSGSIZE 16364
EMSGSIZE 8166
EMSGSIZE 8146
EMSGSIZE 4464
EMSGSIZE 4444
EMSGSIZE 4352
EMSGSIZE 4332
EMSGSIZE 2048
EMSGSIZE 2028
EMSGSIZE 2002
EMSGSIZE 1982
EMSGSIZE 1536
EMSGSIZE 1516
outgoing MTU = 1500
1


here is the oddity:

36 bytes from 127.0.0.1 to 127.0.0.1: icmp type 3 (Dest Unreachable)
code 3
4: x4500005a
8: xc0c50000
12: x1e110000
16: x7f000001
20: x7f000001
24: xf9dd6dc4
28: x00320000
32: x00000000
.....etc etc etc...

repeats this about 20 times , then returns the expected response:

apadb255n2-private (192.168.248.145) 36 bytes to 192.168.248.144 0 ms
0 ms



[ROOT:apadb255n1]/# netstat -r|grep pri
192.168.248.0 apadb255n1-privat UHSb 0 0 en1 -
- =>
192.168.248/24 apadb255n1-privat U 201 91357459 en1 -
-
apadb255n1-priva localhost UGHS 53 39861 lo0 -
-
192.168.248.255 apadb255n1-privat UHSb 0 4 en1 -
-