> we have a customer server with several public IP addresses. Within
> about 2 months we now had for the second time the problem that the
> server did not send e-mails from its eth0 address, but from the
> eth0:0 address.
Yeah, this is indeed problematic. We witnessed similar problems with
callbacks to NewLinQ no longer coming from eth0, but from random other
bound IPv4 IPs, which caused license related issues.
> I have the feeling that the problem has occurred since the
> introduction of IPv6 compatibility. Does anyone have any idea
> what could be the cause of the problem?
I don't have a solid idea yet what the cause might be. When I did the
IPv6 integration I noticed many support topics being raised regarding
RHEL6 and RHEL7 (and clones) that (when IPv6 is being used) outgoing
calls will not always be from the primary IPv6 IP.
However: This shouldn't affect IPv4. But yet we see that this problem
has popped up as well for IPv4.
When the problem happens next time around, could you please send me the
ip -o addr show
ip -o route show
That shows all bound IPv4 IPs (first command) and all IPv4 routes. You
also might want to save a copy of that while it's in "good working
condition" and compare it with the state of things when it's suddenly
using a different IP for outgoing connections.
With best regards
Blueonyx mailing list