1500 is too big with gre if you try to lower the 9000 by 4k it may work just a suggestion. I have seen several issues with the 1500 between instances performance is crappy a lot of frags
Inviato da iPhone () > Il giorno Jul 3, 2014, alle ore 19:13, Clint Dilks <cli...@waikato.ac.nz> ha > scritto: > > Hi, > > I have a three node Icehouse Cluster running on CentOS 6.5 ( using RDO yum > repostories but not installed via packstack ). We are using GRE Tunnels, so > experience the issues where some packets are truncated because they exceed > the 1500 mtu size. > > From what I have read most people are reducing the size of the guests MTU to > work around the issue. We thought another another approach would increasing > the MTU size available. > > So for the appropriate ports we have set the MTU to 9000 on the switch. > > If we change the MTU on all nodes to 9000, then all pre-existing instance > appear to work as expected (both before and after rebooting one or all > nodes). But launching a new instance is always stuck at scheduling. Nothing > is being recorded in /var/log/nova/scheduler.log on the controller. > > If we revert all nodes back to MTU 1500 all starts working again. > > Does anyone have suggestions as to what I am missing ? > > !DSPAM:1,53b60f8b127921259016594! > _______________________________________________ > Mailing list: http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack > Post to : openstack@lists.openstack.org > Unsubscribe : http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack > > > !DSPAM:1,53b60f8b127921259016594! _______________________________________________ Mailing list: http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack Post to : openstack@lists.openstack.org Unsubscribe : http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack
