Daniel Anderson wrote:
Three things to check:
Thanks for the ideas.
1. Look for collisions on the network ports (we're looking for symptoms
of a mis-matched duplex on your network)
solbox1# netstat -ni -I skge0
Name Mtu Net/Dest Address Ipkts Ierrs Opkts Oerrs Collis Queue
skge0 1500 10.0.0.0 10.0.0.1 1640325 0 3560947 0 0 0
fc6box1# netstat -ni
Kernel Interface table
Iface MTU Met RX-OK RX-ERR RX-DRP RX-OVR TX-OK TX-ERR TX-DRP TX-OVR
Flg
eth0 1500 0 7400218 1 0 0 4314345 0 0 0
BMRU
lo 16436 0 161876 0 0 0 161876 0 0 0
LRU
These two are connected by a crossover cable (the singe RX-ERR is old)
2. Run a 'netstat -an | grep SENT' on both ends during the 10 minute
wait (maybe in minute 1 or 2)
No *SENT* only ESTABLISHED:
ssh -X 10.0.0.30
sol (client):
Local Address Remote Address Swind Send-Q Rwind Recv-Q State
10.0.0.1.35651 10.0.0.30.22 18480 6995 49640 0 ESTABLISHED
(vanishes immediately at timeout. I didn't (manage to) see any sort of WAIT)
fc6 (server):
Proto Recv-Q Send-Q Local Address Foreign Address
State
tcp 0 0 10.0.0.30:22 10.0.0.1:35651
ESTABLISHED
(stays at least 30 min after the timeout)
Note the "6995" (Send-Q) at the client side. I get this repeatedly whith ssh -X
10.0.0.30,
but the interactive 'ssh -x 10.0.0.30' sessions (like the one i use to do
netstat on fc6)
have a 0 there. Does that ring any bell?
I have only default tcp-stack settings. Should I play with Solaris ndd |
/etc/system ?
3. Name resolution (IP to name and name to IP) on both ends, do them
each a couple times, make sure you are seeing what you expect
No problems; symetric 1:1 name<=>addr, client&server and no timeouts. The
Solaris box is
the nameserver. Even tried bombarding:
# repeat 20 host 10.0.0.30
The other solaris/FC6 network has a similar config, but not identical, although
I
configured both and most likely have configuered them with the same error, if I
could
figure it out -- or someone could figure it out for me :)
This solbox has nge on the Internet-side (waist of GE, but any way) and skge0
on the
inside (skge0 hasn't been bulletproof, giving sporadic PCI-errors that only
reboot fixes)
The other (more remote) has nge0 on both sides, and has been stable (with the
exception of
this problem), and has a cheap soho GE-switch on the inside.
All FC6 boxes have identical HW with 100T e100 eth0.
The Solarisboxes are "headless" 24*7 home-server (NFS/ZFS,SAMBA,FW/GW) while
FC6 are desktops.
I've moved big files like DVD-images over NFS at the expected speed of ~80Mbps.
--
Med vennlig hilsen / Regards
Basefarm AS
Pål Baltzersen
pb.vcf
Description: Vcard
