Mike Gerdts wrote:
On 5/17/07, Garrett D'Amore <[EMAIL PROTECTED]> wrote:
1468 is due to IP and ICMP header costs. 1468 is the maximum ICMP ECHO
payload. (That's why its different from 1500.)
But the traffic I really cared about (NIS transfer, presumably from
initgroups or similar) was not passing. Once I lowered the MTU hung
login sessions became "unhung".
Yes, these have the same MTU related issues. My comment about 1468 was
really to explain why the failure point wasn't at 1496 vs 1500. Other
protocols will have the same problem.
Could it be that because I had the LDOM configured to use an interface
instance number greater than 1000 that it made false assumptions about
the MTU somewhere in the path? Perhaps I should have directed the
VLAN 49 traffic from e1000g49000 through vnet49. I guess I now have
something to try...
I don't know anything about vnets, but it does suggest to me that one
side or the other isn't configured or or isn't able to deal with the
extra 4 bytes of VLAN header.
-- Garrett
_______________________________________________
networking-discuss mailing list
[email protected]
