From afbf9108d893ede0d147da2afe16c9e6d4bc76d4 Mon Sep 17 00:00:00 2001
From: "James E. Blair" <jeblair@redhat.com>
Date: Thu, 29 Nov 2018 15:35:24 -0800
Subject: [PATCH] OpenStack: store ZK records for launch error nodes

If we get an error on create server, we currently leak the instance
because we don't store the external id of the instance in ZK.  It
should eventually be deleted since it's a leaked instance, but
we try to keep track of as much as possible.  OpenStackSDK can
often return the external id to us in these cases, so handle that
case and store the external id on a ZK record so that the instance
is correctly accounted for.

Change-Id: I7ec448e9a7cf6cd01903bf7b5bf4b07a1c143fb8
---
 nodepool/driver/openstack/handler.py | 40 +++++++++++++++++-----------
 1 file changed, 24 insertions(+), 16 deletions(-)

diff --git a/nodepool/driver/openstack/handler.py b/nodepool/driver/openstack/handler.py
index 47123923c..c52d45849 100644
--- a/nodepool/driver/openstack/handler.py
+++ b/nodepool/driver/openstack/handler.py
@@ -18,6 +18,7 @@ import pprint
 import random
 
 from kazoo import exceptions as kze
+import openstack
 
 from nodepool import exceptions
 from nodepool import nodeutils as utils
@@ -122,22 +123,29 @@ class OpenStackNodeLauncher(NodeLauncher):
         # because that isn't available in ZooKeeper until after the server is
         # active, which could cause a race in leak detection.
 
-        server = self.handler.manager.createServer(
-            hostname,
-            image=image_external,
-            min_ram=self.label.min_ram,
-            flavor_name=self.label.flavor_name,
-            key_name=self.label.key_name,
-            az=self.node.az,
-            config_drive=config_drive,
-            nodepool_node_id=self.node.id,
-            nodepool_node_label=self.node.type[0],
-            nodepool_image_name=image_name,
-            networks=self.pool.networks,
-            security_groups=self.pool.security_groups,
-            boot_from_volume=self.label.boot_from_volume,
-            volume_size=self.label.volume_size,
-            instance_properties=self.label.instance_properties)
+        try:
+            server = self.handler.manager.createServer(
+                hostname,
+                image=image_external,
+                min_ram=self.label.min_ram,
+                flavor_name=self.label.flavor_name,
+                key_name=self.label.key_name,
+                az=self.node.az,
+                config_drive=config_drive,
+                nodepool_node_id=self.node.id,
+                nodepool_node_label=self.node.type[0],
+                nodepool_image_name=image_name,
+                networks=self.pool.networks,
+                security_groups=self.pool.security_groups,
+                boot_from_volume=self.label.boot_from_volume,
+                volume_size=self.label.volume_size,
+                instance_properties=self.label.instance_properties)
+        except openstack.cloud.exc.OpenStackCloudCreateException as e:
+            if e.resource_id:
+                self.node.external_id = e.resource_id
+                # The outer exception handler will handle storing the
+                # node immediately after this.
+            raise
 
         self.node.external_id = server.id
         self.node.hostname = hostname