We've had an 8.5 cluster running with 6 stores for a few years and are in the final stages of building a replacement 8.7 cluster. In both, I built them in the prescribed order (ldap, mta(s), store(s)) and they work. After getting them setup and running (a few years after in the case of the 8.5 cluster), I added another mail store. After doing so, the webmail client started getting a timeout because upstream server is unreachable.
I'm not sure what would cause adding a store in the initial setup to be different from adding one later (especially in the 8.7 case, I took notes and created them in the same way).
It looks like the proxy should be connecting to the upstreams on port 8443 - the stores *are* listening on 8443 and zimbraReverseProxySSLToUpstreamEnabled: TRUE
is set.
After running
sudo -u zimbra /opt/zimbra/libexec/zmproxyconfig -e -w -H zmail04-mta01.peak.org -x redirect
(on each mta using each one's name accordingly)
one of the mtas is working right, but the other two are still getting the upstream timeout.