So we are using nginx from a few builds back as upgrading consistently
has proven difficult for our ops team to keep up with where I work. I
think we’re on 1.0.5.
We have some hosts setup to handle subrequests using proxy_pass
directive. When the DNS record is changed, and the host command on
the machine is reporting a new host ip in the dns result. That host
does not see traffic like it should.
We have always attributed this to caching in dns and we perform “nginx
reconfigure” to reparse conf and re-read dns. This isn’t working for
us though and i’m not sure why. We did upgrade to 1.0.5 relatively
recently and i’m wondering if we inherited a bug. Two things are odd
When we pulled the host ip out of the dns, and performed nginx
reconfigure, the traffic to the machine ceased. But now that our work
is done, and we’re trying to put it back into rotation, nginx
reconfigure does not seem to be working.
Doing a restart is harder, as we have to drain the machine in
question. But is there any known behavior that might be the root
cause of this? Far as I’m aware the TTL on the DNS is 5m. But its
been hours since the change.