Compatiblity with nginx resolve

[splitice]

Is it just my config or is this incompatible with the resolve parameter on upstream servers in nginx 1.27.3?

I havent got a precise resolution yet but it seems to happen if the address of the resolved server changes. The result is that connections hang and then eventually 504.

[RekGRpth]

full reproduction needed

[splitice]

@RekGRpth I'm working on it. Unfortunately its being stubborn and only happening in our staging environment and not my local environment and the builds take forever.

I'm pretty sure it only happens when the upstream server changes IP.

Config that replicates:

    upstream api {
        server api-backend.blah:8000 resolve max_fails=100 max_conns=3;
        keepalive 2;
        queue 512 timeout=15s;
        zone api 64k;
    }

[splitice]

It looks like the specific update pattern of a DNS service in kubernetes has something to do with it.

Potentially something to do with a service going from one or more servers, to zero (while starting) to one or more again (after availability checks). Which would explain replication in our small staging environment.

[splitice]

#7

This seems to be working so far.