ACS 4.4.1 / Router VM (4.4) default gateway "sometimes" missing

[Stephan Seitz <s....@heinlein-support.de>]

Hi,

we're currently running a test-installallation of ACS 4.4.1 w/ XenServer
6.2. The Zone is setup as "Advanced Zone" with VLAN as isolation method.
Configured are "Shared IP"-networks as well as isolated ones.
Sometimes, if a new VM is deployed with an additional, new isolated
network, the corresponding router-VM doesn't set it's own default route.
Working solutions are:
- ssh to the router-VM via link-local and "ip route add default via
N.N.N.N dev eth2"
- destroy the router-VM and get it re-created
- migrate the router-VM a few times to other Hosts (doesn't depend on
which of the available Hosts)

Maybe the Problem is selfmade, but I'm quite unsure: After our initial
setup, we've decided to give a /25 to the Shared-IP Network. A few days
later we split that into 2 /26. The changes to the existing /23 have
been made with all nodes and management down inside the mysql DB.
Existing System and Router VM were destroyed and re-created.
I've double-checked every change and found it 100% working. Due to
logfile-analysis I can't imagine that change responsible.

Since Instances could connect one NIC shared and one NIC isolated, we've
set "network.disable.rpfilter" to "false". Switching it back to "true"
doesn't solve the problem of "sometimes no default route".

Could someone please shed some light, where to look further?


Thanks in advance!

- Stephan