GitLab

The pool utilization for the subnets presented in the dashboard is now
shown together with a progress bar.

 - this affected build on docker on ubuntu 20.04

The embedded exporter registered the named exporter, a version
exporter, and a process exporter, but only unregistered the named
exporter.

Unregister also the process exporter. However, unregistering the
version exporter results in a segmentation fault so that does
perhaps not need to be unregistered?

The groups initialization was moved to the ServerDataService next to the
call that fetches and caches app stats.

Now server data service subscribes to loging/logout changes.
So when user re-logins then server data is automatically refreshed.

This reverts commit bf98a830.

When the user is re-logging to Stork after session expiration the user
is taken to the dashboard page. The page is now correctly displayed.
Previously it would be blank.

The traffic generators now send authentication requests prior to getting
the subnets/servers information. The generators were not functional due
to getting HTTP Error 403 after authentication fixes applied with #119.

Avoid sending unauthorized requests in login page by a non logged in user.

As a result of review.

When the REST API returns error 403 the user is redirected to the 403
page.

All REST calls are guarded by security token by default. It is selectively
disabled for some calls which are required to create new session.

- replaced DefaultRegistry with our own registry in Kea and BIND 9
  exporters
- updated Prometheus library from 1.4.1 to 1.7.0 to use promauto.With
  API

Backed off the chanege which previously added an overview panel to the app
view, per review comments.

Per review comments.

The buttons leading to DHCP subnets etc. are hidden if the daemon is not
active, per review comments.

The message about not monitored daemon is now displayed as warning rather
than information, per review comments.

When clicking the refresh button, the app is now reloaded and refreshed
properly in the UI.

The communication errors with Stork Agent and Kea daemons are now displayed
in the dashboard and apps view.

The error logging has been improved in the areas where the server fails
to communicate with the agent, Kea CA and daemons.

If there is no communication with the daemon, this is shown on the
dashboard with the times icon and the tooltips contain some information
about the problem, e.g. whether it is the Stork agent, Kea CA or a daemon
that is not responding.

The agent, CA and daemon communication error count is now returned over
the REST API.

Added new convenience function which returns statistics for the given
agent selected using address and port pair.

The new data structures hold statistics about the communication failures
with the monitored agents as well as the Kea and Bind9 apps accessible
via those agents.

Instead log as info.