On Tue, Aug 23, 2022 at 10:41 AM <markeczzz(a)gmail.com> wrote:
Found the problem..
I did tcpdump and found that some DNS requests didn't get responeses.
One of upstream DNS servers didn't have TCP enabled for DNS requests so requests that
were sent to that server were unanswered, thats why I had Engine restarts at random times.
Thanks for the update!
Since I don't control upstream DNS servers (my ovirt DNS server is Gateway that just
resends reqests to upstream) .
A tcp-level gateway? It might be not that hard to replace it with a
real caching DNS server. Up to you, of course.
I am thinking that maybe ping would be better in my case for liveness
check..
The reason we changed from ping (the only test we had some time ago)
to dns, and from dns udp to tcp, is that many routers/switches try
harder to handle tcp, under load - they assume that udp is cheaper to
lose. So I'd personally consider the tcp test, unless you have other
means to make sure ping is reliable, or you deliberately want the
engine to move to some other host if ping fails often enough.
Best regards,
--
Didi