With the changes suggested I was able to talk to UML instances on multiple hosts. Thanks for your support.
But I have new issue now. I have an 2 UML instances running on 2 hosts. My application runs on both UML instances and syncs huge data. The data syncup wasn't successfull. I see the following error message on the host machine console: Packet length errorsize 4094 expected size 18178 I tried to bring up my app on host machine and the syncup is perfect. I just wanted to check whether this is caused by vde switch/tap interface(MTU) overhead. Any ideas? Regards, Sirisha On Wed, May 5, 2010 at 11:51 AM, Daniele Lacamera <danie...@users.sourceforge.net> wrote: > Sirisha Alla wrote: >> >> Hi, >> > > Hello. >> >> I was trying to use vde_switch utility and form a virtual network for >> UML instances running on multiple hosts. >> >> Here is my configuration: >> >> Two linux machines: (host1)10.130.130.180, (host2)10.130.131.90 >> >> On host1 did the following: >> >> 1) created tap0 and assigned 192.168.0.254 >> 2) ran vde switch with the following commands: vde_switch -tap tap0 >> -daemon -s /tmp/switch -unix /tmp/switch/ctl >> >> > > you are duplicating the same option twice, that should not happen. > Either use -unix or -s, ensure that /tmp/switch is not already there. > From now on, I will assume your switch command line is just "vde_switch > -tap tap0 -daemon -s /tmp/switch" >> >> On host2 did the following: >> >> 1) created tap0 with192.168.0.100 ip address >> 2) ran vde switch with the following commands: vde_switch -tap tap0 >> -daemon -s /tmp/switch -unix /tmp/switch/ctl >> >> > > same mistake here ^^ > >> - Created vde cable using dpipe vde_plug /tmp/switch/ctl = ssh host2 >> vde_plug /tmp/switch/ctl on host1 and dpipe vde_plug /tmp/switch/ctl = >> ssh host1 vde_plug /tmp/switch/ctl on host2 >> > > Do you use two cables to connect two switches each other in real life > :)? Every dpipe is bidirectional, so you just need one cable with two > plugs. Try just "dpipe vde_plug /tmp/switch = ssh host2 vde_plug > /tmp/switch" on host1, and be careful to use the same path you passed in > switch creation ('ctl' is not needed, vde plugs detect control port > automatically). Also, if you used tap on both sides, you will be able to > ping the machines on the vde network. > Simply pinging 192.168.0.100 from host1 should work at this point, and > should allow you to test your virtual network on your real hosts, before > actually starting the two guest machines. > > Also, consider using an alternative cable: vde_cryptcab for better > network performance, or udp netcat for best if authentication and data > confidentiality are not in your primary goals, (i.e. your real network > is your LAN). Refer to manuals or find a tutorial on > http://wiki.virtualsquare.org/. > > As last optimization, to avoid IP fragments, shrink your tap and UML eth > MTUs to something (50 Bytes) less than the real network ones, to allow > a cleaner encapsulation of 'virtual' ethernet frames into 'real' > transport payloads. > > [uml stuff following] >> >> I will be happy to provide suggested logs if additional information is >> required. My requirement is that both 192.168.0.1 and 192.168.0.3 >> should see each other. >> >> Regards, >> Sirisha >> > > HTH > ------------------------------------------------------------------------------ _______________________________________________ vde-users mailing list vde-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/vde-users