ipv6 support

projects/elife.yaml

@@ -94,22 +94,40 @@ defaults:
             # optional: specify `ports` to be opened
             security-group: {}
         region: us-east-1
-        vpc-id: vpc-78a2071d  # vpc-id + subnet-id are peculiar to AWS account + region
+        vpc-id: vpc-78a2071d
 
-        subnet-id: subnet-1d4eb46a # elife-public-subnet, us-east-1d
-        subnet-cidr: '10.0.2.0/24'
+        # project supports IPv6.
+        use-ipv6: false # create DNS AAAA records, add IPv6 security group CIDRs
+
+        subnet-id: subnet-1d4eb46a # elife-public-subnet
+        subnet-cidr: '10.0.2.0/24' # *internal* subnet-cidr, always ipv4, used with RDS 
         availability-zone: us-east-1d
 
-        redundant-subnet-id: subnet-7a31dd46 # elife-public-subnet-2, us-east-1e
+        # ipv6 alternate
+        #subnet-id: subnet-0866e00eb5b9e27d4 # ipv6-elife-public-subnet
+        #subnet-cidr: '10.0.15.0/24' # contradiction? no! subnet is not *pure* ipv6. internally it uses ipv4.
+        #availability-zone: us-east-1d
+
+        redundant-subnet-id: subnet-7a31dd46 # elife-public-subnet-2
         redundant-subnet-cidr: '10.0.3.0/24'
         redundant-availability-zone: us-east-1e
 
+        # ipv6 alternate
+        #redundant-subnet-id: subnet-02000374730149bf6 # ipv6-elife-public-subnet-2
+        #redundant-subnet-cidr: '10.0.16.0/24'
+        #redundant-availability-zone: us-east-1e
+
         # lsh@2022-06-27: t3.* instance types not available in us-east-1e
         # - https://aws.amazon.com/premiumsupport/knowledge-center/ec2-instance-type-not-supported-az-error/
-        redundant-subnet-id-2: subnet-2116727b # elife-public-subnet-3, us-east-1a
+        redundant-subnet-id-2: subnet-2116727b # elife-public-subnet-3
         redundant-subnet-cidr-2: '10.0.10.0/24'
         redundant-availability-zone-2: us-east-1a
 
+        # ipv6 alternate
+        #redundant-subnet-id-2: subnet-03a5217409cbe6633 # ipv6-elife-public-subnet-3
+        #redundant-subnet-cidr-2: 10.0.17.0/24
+        #redundant-availability-zone-2: us-east-1a
+
         rds:
             multi-az: false
             engine: postgres # or 'MySQL'
@@ -445,6 +463,14 @@ basebox:
             ami: ami-027a754129abb5386 # focal, build 20231208, hvm:ebs-ssd
             ports:
                 - 22
+    aws-alt:
+        ipv6test:
+            use-ipv6: true
+            subnet-id: subnet-0866e00eb5b9e27d4 # ipv6-elife-public-subnet
+            subnet-cidr: '10.0.15.0/24'
+            ec2:
+                ami: ami-08d0e948516bdfee8 # GENERATED created from basebox--2004
+
     vagrant: {}
 
 heavybox:
@@ -1490,8 +1516,8 @@ search:
     vagrant:
         ram: 4096
         ports:
-            1244: 80
-            1245: 8080
+            1244: 80   # webserver
+            1245: 8080 # api-dummy
             9921: 9201 # opensearch
 
 recommendations:

src/buildercore/cfngen.py

@@ -332,6 +332,8 @@ def build_context_aws(pdata, context):
         'account-id',
         'vpc-id',
 
+        'use-ipv6',
+
         'subnet-id',
         'subnet-cidr',
         'availability-zone',
@@ -1094,7 +1096,7 @@ def _current_cloudformation_template(stackname):
         raise
 
 def download_cloudformation_template(stackname):
-    cloudformation.write_template(stackname, json.dumps(_current_cloudformation_template(stackname)))
+    cloudformation.write_template(stackname, json.dumps(_current_cloudformation_template(stackname), indent=4))
 
 def regenerate_stack(stackname, **more_context):
     current_context = context_handler.load_context(stackname)

src/buildercore/core.py

@@ -245,6 +245,28 @@ def find_ec2_instances(stackname, state='running', node_ids=None, allow_empty=Fa
         raise NoRunningInstancesError("found no running ec2 instances for %r. The stack nodes may have been stopped, but here we were requiring them to be running" % stackname)
     return ec2_instances
 
+# prefer this over `find_ec2_instances`
+def ec2_data(stackname, state=None):
+    """returns a list of raw boto3 EC2.Instance data for ec2 instances attached to given `stackname`.
+    does not filter by state by default.
+    does not enforce single instance checking."""
+    try:
+        ec2_instances = find_ec2_instances(stackname, state=state, allow_empty=True)
+        return [ec2.meta.data for ec2 in ec2_instances]
+    except Exception:
+        LOG.exception('unhandled exception attempting to discover more information about this instance. Instance may not exist yet.')
+        raise
+
+def pick_ip_address(ec2_data):
+    """returns a public IP address, preferring ipv4 when available over ipv6.
+    for use with the ec2 instance data returned by `ec2_data`."""
+    return ec2_data.get('PublicIpAddress') or ec2_data.get('Ipv6Address')
+
+def pick_ip_address_obj(ec2_obj):
+    """returns a public IP address, preferring ipv4 when available over ipv6.
+    for use with the ec2 instance objects returned by `find_ec2_instances`."""
+    return ec2_obj.public_ip_address or ec2_obj.ipv6_address
+
 #
 #
 #
@@ -344,7 +366,7 @@ def stack_conn(stackname, username=config.DEPLOY_USER, node=None, **kwargs):
     node and ensure(utils.isint(node) and int(node) > 0, "given node must be an integer and greater than zero")
     didx = int(node) - 1 if node else 0 # decrement to a zero-based value
     data = data[didx] # data is ordered by node
-    public_ip = data['PublicIpAddress']
+    public_ip = pick_ip_address(data)
     params = _ec2_connection_params(stackname, username, host_string=public_ip)
 
     with settings(**params):
@@ -356,7 +378,7 @@ class NoPublicIpsError(Exception):
 def all_node_params(stackname):
     "returns a map of node data"
     data = ec2_data(stackname, state='running')
-    public_ips = {ec2['InstanceId']: ec2.get('PublicIpAddress') for ec2 in data}
+    public_ips = {ec2['InstanceId']: pick_ip_address(ec2) for ec2 in data}
     nodes = {
         ec2['InstanceId']: int(tags2dict(ec2['Tags'])['Node'])
         if 'Node' in tags2dict(ec2['Tags'])
@@ -385,7 +407,7 @@ def stack_all_ec2_nodes(stackname, workfn, username=config.DEPLOY_USER, concurre
 
     data = ec2_data(stackname, state='running')
     # TODO: reuse all_node_params?
-    public_ips = {ec2['InstanceId']: ec2.get('PublicIpAddress') for ec2 in data}
+    public_ips = {ec2['InstanceId']: pick_ip_address(ec2) for ec2 in data}
     nodes = {ec2['InstanceId']: int(tags2dict(ec2['Tags'])['Node']) if 'Node' in tags2dict(ec2['Tags']) else 1 for ec2 in data}
     if node:
         nodes = {k: v for k, v in nodes.items() if v == int(node)}
@@ -586,16 +608,6 @@ def describe_stack(stackname, allow_missing=False):
 class NoRunningInstancesError(Exception):
     pass
 
-def ec2_data(stackname, state=None):
-    """returns a list of raw boto3 EC2.Instance data for ec2 instances attached to given `stackname`.
-    does not filter by state by default.
-    does not enforce single instance checking."""
-    try:
-        ec2_instances = find_ec2_instances(stackname, state=state, allow_empty=True)
-        return [ec2.meta.data for ec2 in ec2_instances]
-    except Exception:
-        LOG.exception('unhandled exception attempting to discover more information about this instance. Instance may not exist yet.')
-        raise
 
 
 # DO NOT CACHE: function is used in polling

src/buildercore/lifecycle.py

@@ -371,7 +371,7 @@ def _wait_for_running_nodes(stackname):
     if context['ec2'].get('dns-external-primary'):
         primary = 1
         primary_hostname = context['ext_node_hostname'] % primary
-        primary_ip_address = nodes[0].public_ip_address
+        primary_ip_address = core.pick_ip_address_obj(nodes[0])
         LOG.info("External primary full hostname: %s", primary_hostname)
         _update_dns_a_record(context['domain'], primary_hostname, primary_ip_address)
 
@@ -384,7 +384,7 @@ def _wait_for_running_nodes(stackname):
     LOG.info("External full hostname: %s", context['full_hostname'])
     if context['full_hostname']:
         for node in nodes:
-            _update_dns_a_record(context['domain'], context['full_hostname'], node.public_ip_address)
+            _update_dns_a_record(context['domain'], context['full_hostname'], core.pick_ip_address_obj(node))
 
 def _delete_dns_a_record(zone_name, name):
     """deletes a Route53 DNS 'A' record `name` in hosted zone `zone_name`.