Azure: eject the provisioning iso before reporting ready (#861)

Due to hyper-v implementations, iso ejection is more efficient if performed
from within the guest. The code will attempt to perform a best-effort ejection.
Failure during ejection will not prevent reporting ready from happening. If iso
ejection is successful, later iso ejection from the platform will be a no-op.
In the event the iso ejection from the guest fails, iso ejection will still happen at
the platform level.

2 files changed, 39 insertions, 6 deletions

diff --git a/cloudinit/sources/DataSourceAzure.py b/cloudinit/sources/DataSourceAzure.py
index 8c7fb021..eac6405a 100755
--- a/cloudinit/sources/DataSourceAzure.py
+++ b/cloudinit/sources/DataSourceAzure.py
@@ -332,6 +332,7 @@ class DataSourceAzure(sources.DataSource):
     dsname = 'Azure'
     _negotiated = False
     _metadata_imds = sources.UNSET
+    _ci_pkl_version = 1
 
     def __init__(self, sys_cfg, distro, paths):
         sources.DataSource.__init__(self, sys_cfg, distro, paths)
@@ -346,8 +347,13 @@ class DataSourceAzure(sources.DataSource):
         # Regenerate network config new_instance boot and every boot
         self.update_events['network'].add(EventType.BOOT)
         self._ephemeral_dhcp_ctx = None
-
         self.failed_desired_api_version = False
+        self.iso_dev = None
+
+    def _unpickle(self, ci_pkl_version: int) -> None:
+        super()._unpickle(ci_pkl_version)
+        if "iso_dev" not in self.__dict__:
+            self.iso_dev = None
 
     def __str__(self):
         root = sources.DataSource.__str__(self)
@@ -459,6 +465,13 @@ class DataSourceAzure(sources.DataSource):
                     '%s was not mountable' % cdev, logger_func=LOG.warning)
                 continue
 
+            report_diagnostic_event("Found provisioning metadata in %s" % cdev,
+                                    logger_func=LOG.debug)
+
+            # save the iso device for ejection before reporting ready
+            if cdev.startswith("/dev"):
+                self.iso_dev = cdev
+
             perform_reprovision = reprovision or self._should_reprovision(ret)
             perform_reprovision_after_nic_attach = (
                 reprovision_after_nic_attach or
@@ -1226,7 +1239,9 @@ class DataSourceAzure(sources.DataSource):
         @return: The success status of sending the ready signal.
         """
         try:
-            get_metadata_from_fabric(None, lease['unknown-245'])
+            get_metadata_from_fabric(fallback_lease_file=None,
+                                     dhcp_opts=lease['unknown-245'],
+                                     iso_dev=self.iso_dev)
             return True
         except Exception as e:
             report_diagnostic_event(
@@ -1332,7 +1347,8 @@ class DataSourceAzure(sources.DataSource):
         metadata_func = partial(get_metadata_from_fabric,
                                 fallback_lease_file=self.
                                 dhclient_lease_file,
-                                pubkey_info=pubkey_info)
+                                pubkey_info=pubkey_info,
+                                iso_dev=self.iso_dev)
 
         LOG.debug("negotiating with fabric via agent command %s",
                   self.ds_cfg['agent_command'])
diff --git a/cloudinit/sources/helpers/azure.py b/cloudinit/sources/helpers/azure.py
index 03e7156b..ad476076 100755
--- a/cloudinit/sources/helpers/azure.py
+++ b/cloudinit/sources/helpers/azure.py
@@ -865,7 +865,19 @@ class WALinuxAgentShim:
         return endpoint_ip_address
 
     @azure_ds_telemetry_reporter
-    def register_with_azure_and_fetch_data(self, pubkey_info=None) -> dict:
+    def eject_iso(self, iso_dev) -> None:
+        try:
+            LOG.debug("Ejecting the provisioning iso")
+            subp.subp(['eject', iso_dev])
+        except Exception as e:
+            report_diagnostic_event(
+                "Failed ejecting the provisioning iso: %s" % e,
+                logger_func=LOG.debug)
+
+    @azure_ds_telemetry_reporter
+    def register_with_azure_and_fetch_data(self,
+                                           pubkey_info=None,
+                                           iso_dev=None) -> dict:
         """Gets the VM's GoalState from Azure, uses the GoalState information
         to report ready/send the ready signal/provisioning complete signal to
         Azure, and then uses pubkey_info to filter and obtain the user's
@@ -891,6 +903,10 @@ class WALinuxAgentShim:
             ssh_keys = self._get_user_pubkeys(goal_state, pubkey_info)
         health_reporter = GoalStateHealthReporter(
             goal_state, self.azure_endpoint_client, self.endpoint)
+
+        if iso_dev is not None:
+            self.eject_iso(iso_dev)
+
         health_reporter.send_ready_signal()
         return {'public-keys': ssh_keys}
 
@@ -1046,11 +1062,12 @@ class WALinuxAgentShim:
 
 @azure_ds_telemetry_reporter
 def get_metadata_from_fabric(fallback_lease_file=None, dhcp_opts=None,
-                             pubkey_info=None):
+                             pubkey_info=None, iso_dev=None):
     shim = WALinuxAgentShim(fallback_lease_file=fallback_lease_file,
                             dhcp_options=dhcp_opts)
     try:
-        return shim.register_with_azure_and_fetch_data(pubkey_info=pubkey_info)
+        return shim.register_with_azure_and_fetch_data(
+            pubkey_info=pubkey_info, iso_dev=iso_dev)
     finally:
         shim.clean_up()