Remove any db queries during physical restore

For physical restore, we cannot assume that the database nodes are up and available before kicking off the restore. It is very likely the restore is required because the cluster is in a bad shape.

PBM agent currently relies on the local mongod process (and also config server replica set in the case of sharded clusters) and won’t start if it cannot communicate with it/them.

We should still be able to start a restore by reading any data required directly from the configured backup location. Storage-based communication is enough for physical restore since it does not matter which node is primary.

This will require saving the following information as part of backup metadata:

For replica sets:

• replica set name

• FQDN and port of each member

• type of each member (data node or arbiter)

For sharded clusters:

• config server replica set name

• FQDN and port of all members of the config server replica set

• replica set name of each shard

• FQDN and port of all members of each shard

• type of each member (data node or arbiter)