[LIBCLOUD-879] GCE Load balancer suppose that all VM have public ip

• Type: Bug
• Status: Closed
• Priority: Minor
• Resolution: Fixed
• Component/s: LoadBalancer

Environment

GCE MASTER

• Flags: Patch
• Assignee: Unassigned
• Reporter: charles walker
• Created: 22/Nov/16 18:18
• Updated: 25/Jan/18 12:11
• Resolved: 23/Dec/16 14:33

Description

When calling "list_members()" on a GCe load balancer libcloud object I end up with the following stack :

Retrieve 2 Lbs. Here is the list : [<LoadBalancer: id=68XXXXXXXX707, name=europe-lb-forwarding-rule, state=None, ip=10XXXXXXXX.77, port=30012-30012>, <LoadBalancer: id=87688XXXXXXXX64, name=load-balancer-prd-comp-forwarding-rule-2, state=None, ip=10XXXXXXXX93, port=30001-30001>]
--Printing the node obj for DBG : <Node: uuid=9001a5d9d425dc0e5dd1db5352296b08920bde21, name=tec-XXXXXXXX-infra-2ktm, state=RUNNING, public_ips=[], private_ips=['10.XXXXXXXX'], provider=Google Compute Engine ...>
Traceback (most recent call last):
  File "LbTestPy.py", line 43, in <module>
    print ("Members: " +str(aLbs[0].list_members()))
  File "/home/cloud-user/LbTest/src/apache-libcloud/libcloud/loadbalancer/base.py", line 110, in list_members
    return self.driver.balancer_list_members(balancer=self)
  File "/home/cloud-user/LbTest/src/apache-libcloud/libcloud/loadbalancer/drivers/gce.py", line 274, in balancer_list_members
    balancer.extra['targetpool'].nodes]
  File "/home/cloud-user/LbTest/src/apache-libcloud/libcloud/loadbalancer/drivers/gce.py", line 342, in _node_to_member
    member_ip = node.public_ips[0]
IndexError: list index out of range

After some investigations it appears to comes from the method "def _node_to_member(self, node, balancer):" in loadbalancer/drivers/gce.py. Mode precisely this piece of the code :

code1.py

if hasattr(node, 'name'):
	member_id = node.name
	member_ip = node.public_ips[0]
else:
	member_id = node
	member_ip = None

which imply that all VM in the load balancer will have public IP. This is not necessarly the case and thus when it happen the process crash (as you can see in the previous stack where i added some debug log in my libcloud version to print the node where the error occurs).

I would had suggest to use private ip instead of public but I do not want to impact the existing user of libcloud so I was thinking of a simple fix :

code1.py

if hasattr(node, 'name'):
    member_id = node.name
else:
    member_id = node

if (len(node.public_ips) > 0):
    member_ip = node.public_ips[0]
else:
    member_ip = None

Issue Links

• links to: GitHub Pull Request #952