You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Configuring the cluster to use a host/domain name for the http.publish_host setting (useful for reverse proxies) does not work with connection pools. It seems that Elasticsearch itself tries to resolve the value of the setting into an IP address and stores both the name and the address as a publish_address. Querying a cluster set up like this with, e.g., /_nodes, shows that the http_address field has a value like this, for example:
<FQDN>/<IP>:<port>
When using a sniffing connection pool, and once a sniff has been completed, the pool will be reseeded with nodes using the above as the base URL. This means that subsequent calls have the following form:
<http(s)>://<FQDN>/<IP>:<port>/rest/of/url/in/request (with port 80 used implicitly for HTTP requests)
Before trying this, I would have expected the following:
<http(s)>://<FQDN>:<port>/rest/of/url/in/request
Where FQDN would be the value configured for http.publish_host and port the value configured for http.publish_port.