At the moment the ArangoDatabaseFacade unconditionally activates aquireHostList on the ArangoDB driver which is used for active failover. This leads to issues when ArangoDB is deployed in a cluster mode. For example in Kubernetes Spline is supposed to communicate to the the database via a service IP, not directly to pods, but with enabled aquireHostList the driver would fetched the direct pods IPs and use those for subsequent requests (basically contradicting K8s practices).
It creates even more issues to Admin CLI when connecting to the database via an external IP from outside of the cluster. Obtained pod host names or IPs aren't valid outside of the cluster and it breaks the admin app.
The aquireHostList needs to be configurable for the server, and disabled for the admin, as the admin app, with the respect to the way it's used, doesn't really require the DB to be highly available.
At the moment the
ArangoDatabaseFacadeunconditionally activatesaquireHostListon the ArangoDB driver which is used for active failover. This leads to issues when ArangoDB is deployed in a cluster mode. For example in Kubernetes Spline is supposed to communicate to the the database via a service IP, not directly to pods, but with enabledaquireHostListthe driver would fetched the direct pods IPs and use those for subsequent requests (basically contradicting K8s practices).It creates even more issues to Admin CLI when connecting to the database via an external IP from outside of the cluster. Obtained pod host names or IPs aren't valid outside of the cluster and it breaks the admin app.
The
aquireHostListneeds to be configurable for the server, and disabled for the admin, as the admin app, with the respect to the way it's used, doesn't really require the DB to be highly available.