That is correct. Apache will be a single point of failure in the scenario you describe. To eliminate that, you would want to provide HA at the apache tier. This is generally done via a content switch (BigIP, LocalDirector, Cisco CSM, etc), where the switch provides a virtual IP/host and the routes request to the apache tier, according to the configured policy.
You could also achieve via DNS round robin across two apache servers, but the outage detection may take longer due to DNS TTL.
Mike