Do not break the VM list when a VM's OpenNebula template was deleted

The "my virtual machines" page showed "We could not load your VMs due to a
backend connection error" for any user owning a VM whose template had been
removed from OpenNebula. OpenNebulaManager.get_template() converted every
failure, including WrongIdError for a missing template, into
ConnectionRefusedError, and VirtualMachineSerializer.get_configuration()
let that propagate, so one VM with a stale template id took down the whole
list. The only log output was an empty ERROR line, because WrongIdError
has no message.

* Add opennebula_api.exceptions.TemplateNotFoundError and raise it from
  get_template() for a missing template; keep ConnectionRefusedError for
  genuine backend failures and log them with a traceback.
* Add OpenNebulaManager.get_template_names() and resolve_template_id() so
  the serializer fetches the template pool once per list instead of
  constructing an OpenNebulaManager and fetching the pool for every VM.
* get_configuration() now falls back to the template name embedded in the
  VM name ("<template>-<vm id>") when the template is missing or the pool
  cannot be fetched, and logs why.
* Replace str.lstrip('public-') with a real prefix strip (lstrip removes a
  character set, so "public-ubuntu" became "ntu") and also strip
  "ipv6only-".
* ProcessVMSelectionMixin shows "configuration is no longer available"
  instead of an unhandled error when ordering a retired template.
* VirtualMachinesPlanListView logs the exception behind the banner.
* Add mock-based unit tests for the fallback, the UPDATED_TEMPLATES
  mapping, the single pool fetch and the new exception.

The UPDATED_TEMPLATES setting keeps working as an operational override for
mapping retired template ids to their replacements.

hosting/mixins.py

@@ -1,7 +1,10 @@
+from django.contrib import messages
 from django.shortcuts import redirect
 from django.conf import settings
 from django.core.urlresolvers import reverse
+from django.utils.translation import ugettext_lazy as _
 
+from opennebula_api.exceptions import TemplateNotFoundError
 from opennebula_api.serializers import VirtualMachineTemplateSerializer
 from opennebula_api.models import OpenNebulaManager
 
@@ -14,7 +17,15 @@ class ProcessVMSelectionMixin(object):
 
         template_id = int(request.POST.get('vm_template_id'))
         configuration_id = int(request.POST.get('configuration'))
-        template = OpenNebulaManager().get_template(template_id)
+        try:
+            template = OpenNebulaManager().get_template(template_id)
+        except TemplateNotFoundError:
+            messages.error(
+                request,
+                _('The selected configuration is no longer available. '
+                  'Please choose another one.')
+            )
+            return redirect(reverse('hosting:index'))
         data = VirtualMachineTemplateSerializer(template).data
         configuration = HostingPlan.objects.get(id=configuration_id)
 

hosting/views.py

@@ -1557,6 +1557,11 @@ class VirtualMachinesPlanListView(LoginRequiredMixin, ListView):
             serializer = VirtualMachineSerializer(queryset, many=True)
             return serializer.data
         except ConnectionRefusedError:
+            logger.exception(
+                "Could not load VMs of user %s (%s) from OpenNebula" % (
+                    owner.username, owner.email
+                )
+            )
             messages.error(self.request,
                            'We could not load your VMs due to a backend connection \
                 error. Please try again in a few minutes'