On Jan 4, 2005, at 4:02 PM, Alan Cox wrote:
On Maw, 2005-01-04 at 14:20, James Melin wrote:
I'm in an interesting situation..... Our network, hence my linux
guests,
use an MTU size of 1500. As it turns out, our z/os systems use an
MTU size
of 1492. We just started using a z/os HTTP server outside the
firewall to
do reverse proxy inside to our linux guests. When I do not go
through the
z/OS httpd server, things go swimmingly. I suspect I'm getting
retransmit
delays with an 8 byte difference in MTU. I'm not sure under SLES8
where to
change the MTU size for ETH0 to see if making those match will change
the
behaviour we are seeing.
Anyone know where to tweak this particular thing after the fact?
Properly configured internet systems just work. The 1500 MTU packet
will
be fragmented when it hits a point on the network it does not fit.
Hosts
can chose to get an error instead and reduce the sizes themselves
("path
mtu discovery").
If you are getting hangs when talking through your firewall for large
but not small pages it normally means its ICMP filtering for ICMP must
fragment unreachable frames is not correct. Bad performance but no
hangs
would point elsewhere.
All this is true, but it may well indicate a bug in z/OS TCP/IP. How
recent is your z/OS TCP/IP? There was a similar sort of problem on VM
TCP/IP that was fixed several years ago. In any case, try "ifconfig
eth0 mtu 1492" and see if the problem goes away.
Adam
----------------------------------------------------------------------
For LINUX-390 subscribe / signoff / archive access instructions,
send email to [EMAIL PROTECTED] with the message: INFO LINUX-390 or visit
http://www.marist.edu/htbin/wlvindex?LINUX-390
