I'm looking around to try and figure out if I'm hitting a bug/known issue or if I'm doing something out of order, but I can reliably kill the networking on a second XCP 1.0 node if I try to add it to a pool where the master has a bond. My steps have been:
Install XCP on both nodes.
SSH in, create /etc/xensource/xapi_version_override containing the single line "5.6.199".
Restart xapi (/etc/init.d/xapi restart).
Connect to both hosts via XenCenter. At this point the management interface is still eth0.
Create a pool including only xenserver01. Once it is in the pool, I create a bond from eth0 and eth1 (these are PE2950s with only two interfaces, for testing).
So far so good, so I add xenserver02 (you can't a node to a pool if it already has bonds, so I haven't changed anything at all on this node yet other than xapi version).
It adds xenserver02 to the pool, nice green icon and local storage listed. Viewing NICs tab, the bond is created, but shows "unknown" for connectivity, unlike the first pool memeber which shows connected. No VMs can run on xenserver02, and if I go to the console there is no bond0 interface and eth0 is down. Figuring "well, perhaps a reboot to finalize pool membership?", I bounce xenserver02, but when it comes up it is in the same configuration. No bond0, eth0 down, and the local console shows the management network as "<unknown>".
While xenserver01 has an interface named xapi1 (as well as bond0 and both ethX interfaces up), xenserver02 has no bond or xapi interfaces and eth0 isn't configured. Host is now dead with no networking. So, I can fully reinstall XCP and attempt to add to the pool again, but the exact same results.
Anyone else using XCP 1.0 and XenCenter? Are you running into problems similar to this or am I very unlucky? Maybe I've missed a best practices doc on this, but so far nothing.
Thanks for any help/advice,
P.S. Is the license warning in XenCenter harmless, or do "bad things" happen after its 30-day grace period even though these are XCP nodes?
Xen-users mailing list