Blame SOURCES/bz1565670-fence_azure_arm-network-fencing.patch

daee9b
diff -uNr a/fence/agents/azure_arm/fence_azure_arm.py b/fence/agents/azure_arm/fence_azure_arm.py
daee9b
--- a/fence/agents/azure_arm/fence_azure_arm.py	2018-03-09 16:14:51.357065582 +0100
daee9b
+++ b/fence/agents/azure_arm/fence_azure_arm.py	2018-03-13 13:10:56.786200236 +0100
daee9b
@@ -3,19 +3,23 @@
daee9b
 import sys, re, pexpect
daee9b
 import logging
daee9b
 import atexit
daee9b
+import xml.etree.ElementTree as ET
daee9b
 sys.path.append("/usr/share/fence")
daee9b
 from fencing import *
daee9b
-from fencing import fail, fail_usage, EC_TIMED_OUT, run_delay
daee9b
+from fencing import fail_usage, run_command, run_delay
daee9b
+import azure_fence
daee9b
 
daee9b
 #BEGIN_VERSION_GENERATION
daee9b
-RELEASE_VERSION="4.0.25.34-695e-dirty"
daee9b
-BUILD_DATE="(built Wed Jun 28 08:13:44 UTC 2017)"
daee9b
-REDHAT_COPYRIGHT="Copyright (C) Red Hat, Inc. 2004-2010 All rights reserved."
daee9b
+RELEASE_VERSION="Fence agent for Azure Resource Manager"
daee9b
+REDHAT_COPYRIGHT=""
daee9b
+BUILD_DATE=""
daee9b
 #END_VERSION_GENERATION
daee9b
 
daee9b
-def get_nodes_list(compute_client, options):
daee9b
+def get_nodes_list(clients, options):
daee9b
     result = {}
daee9b
-    if compute_client:
daee9b
+
daee9b
+    if clients:
daee9b
+        compute_client = clients[0]
daee9b
         rgName = options["--resourceGroup"]
daee9b
         vms = compute_client.virtual_machines.list(rgName)
daee9b
         try:
daee9b
@@ -26,41 +30,100 @@
daee9b
 
daee9b
     return result
daee9b
 
daee9b
-def get_power_status(compute_client, options):
daee9b
+def check_unfence(clients, options):
daee9b
+    if clients:
daee9b
+        compute_client = clients[0]
daee9b
+        network_client = clients[1]
daee9b
+        rgName = options["--resourceGroup"]
daee9b
+
daee9b
+        try:
daee9b
+            vms = compute_client.virtual_machines.list(rgName)
daee9b
+        except Exception as e:
daee9b
+            fail_usage("Failed: %s" % e)
daee9b
+
daee9b
+        for vm in vms:
daee9b
+            vmName = vm.name
daee9b
+            if azure_fence.get_network_state(compute_client, network_client, rgName, vmName) == "off":
daee9b
+                logging.info("Found fenced node " + vmName)
daee9b
+                # dont return "off" based on network-fencing status
daee9b
+                options.pop("--network-fencing", None)
daee9b
+                options["--plug"] = vmName
daee9b
+                if get_power_status(clients, options) == "off":
daee9b
+                    logging.info("Unfencing " + vmName)
daee9b
+                    options["--network-fencing"] = ""
daee9b
+                    options["--action"] = "on"
daee9b
+                    set_power_status(clients, options)
daee9b
+                    options["--action"] = "monitor"
daee9b
+
daee9b
+def get_power_status(clients, options):
daee9b
+    vmstate = { "running": "on",
daee9b
+                "deallocated": "off",
daee9b
+                "stopped": "off" }
daee9b
     logging.info("getting power status for VM " + options["--plug"])
daee9b
 
daee9b
-    if compute_client:
daee9b
+    if clients:
daee9b
+        compute_client = clients[0]
daee9b
         rgName = options["--resourceGroup"]
daee9b
         vmName = options["--plug"]
daee9b
 
daee9b
+        if "--network-fencing" in options:
daee9b
+            network_client = clients[1]
daee9b
+            netState =  azure_fence.get_network_state(compute_client, network_client, rgName, vmName)
daee9b
+            logging.info("Found network state of VM: " + netState)
daee9b
+
daee9b
+            # return off quickly once network is fenced instead of waiting for vm state to change
daee9b
+            if options["--action"] == "off" and netState == "off":
daee9b
+                logging.info("Network fenced for " + vmName)
daee9b
+                return netState
daee9b
+
daee9b
         powerState = "unknown"
daee9b
         try:
daee9b
             vmStatus = compute_client.virtual_machines.get(rgName, vmName, "instanceView")
daee9b
         except Exception as e:
daee9b
             fail_usage("Failed: %s" % e)
daee9b
+
daee9b
         for status in vmStatus.instance_view.statuses:
daee9b
             if status.code.startswith("PowerState"):
daee9b
-                powerState = status.code
daee9b
+                powerState = status.code.split("/")[1]
daee9b
                 break
daee9b
 
daee9b
-        logging.info("Found power state of VM: " + powerState)
daee9b
-        if powerState == "PowerState/running":
daee9b
+        vmState = vmstate.get(powerState, "unknown")
daee9b
+        logging.info("Found power state of VM: %s (%s)" % (vmState, powerState))
daee9b
+
daee9b
+        if "--network-fencing" in options and netState == "off":
daee9b
+            return "off"
daee9b
+
daee9b
+        if options["--action"] != "on" and vmState != "off":
daee9b
             return "on"
daee9b
 
daee9b
-    return "off"
daee9b
+        if vmState == "on":
daee9b
+            return "on"
daee9b
 
daee9b
-def set_power_status(compute_client, options):
daee9b
+        return "off"
daee9b
+
daee9b
+def set_power_status(clients, options):
daee9b
     logging.info("setting power status for VM " + options["--plug"] + " to " + options["--action"])
daee9b
 
daee9b
-    if compute_client:
daee9b
+    if clients:
daee9b
+        compute_client = clients[0]
daee9b
         rgName = options["--resourceGroup"]
daee9b
         vmName = options["--plug"]
daee9b
 
daee9b
+        if "--network-fencing" in options:
daee9b
+            network_client = clients[1]
daee9b
+
daee9b
+            if (options["--action"]=="off"):
daee9b
+                logging.info("Fencing network for " + vmName)
daee9b
+                azure_fence.set_network_state(compute_client, network_client, rgName, vmName, "block")
daee9b
+            elif (options["--action"]=="on"):
daee9b
+                logging.info("Unfencing network for " + vmName)
daee9b
+                azure_fence.set_network_state(compute_client, network_client, rgName, vmName, "unblock")
daee9b
+
daee9b
         if (options["--action"]=="off"):
daee9b
-            logging.info("Deallocating " + vmName + "in resource group " + rgName)
daee9b
+            logging.info("Deallocating " + vmName + " in resource group " + rgName)
daee9b
             compute_client.virtual_machines.deallocate(rgName, vmName)
daee9b
         elif (options["--action"]=="on"):
daee9b
-            logging.info("Starting " + vmName + "in resource group " + rgName)
daee9b
+            logging.info("Starting " + vmName + " in resource group " + rgName)
daee9b
             compute_client.virtual_machines.start(rgName, vmName)
daee9b
 
daee9b
 
daee9b
@@ -69,8 +132,8 @@
daee9b
         "getopt" : ":",
daee9b
         "longopt" : "resourceGroup",
daee9b
         "help" : "--resourceGroup=[name]         Name of the resource group",
daee9b
-        "shortdesc" : "Name of resource group.",
daee9b
-        "required" : "1",
daee9b
+        "shortdesc" : "Name of resource group. Metadata service is used if the value is not provided.",
daee9b
+        "required" : "0",
daee9b
         "order" : 2
daee9b
     }
daee9b
     all_opt["tenantId"] = {
daee9b
@@ -78,23 +141,56 @@
daee9b
         "longopt" : "tenantId",
daee9b
         "help" : "--tenantId=[name]              Id of the Azure Active Directory tenant",
daee9b
         "shortdesc" : "Id of Azure Active Directory tenant.",
daee9b
-        "required" : "1",
daee9b
+        "required" : "0",
daee9b
         "order" : 3
daee9b
     }
daee9b
     all_opt["subscriptionId"] = {
daee9b
         "getopt" : ":",
daee9b
         "longopt" : "subscriptionId",
daee9b
         "help" : "--subscriptionId=[name]        Id of the Azure subscription",
daee9b
-        "shortdesc" : "Id of the Azure subscription.",
daee9b
-        "required" : "1",
daee9b
+        "shortdesc" : "Id of the Azure subscription. Metadata service is used if the value is not provided.",
daee9b
+        "required" : "0",
daee9b
         "order" : 4
daee9b
     }
daee9b
+    all_opt["network-fencing"] = {
daee9b
+        "getopt" : "",
daee9b
+        "longopt" : "network-fencing",
daee9b
+        "help" : "--network-fencing              Use network fencing. See NOTE-section of\n\
daee9b
+                                  metadata for required Subnet/Network Security\n\
daee9b
+                                  Group configuration.",
daee9b
+        "shortdesc" : "Use network fencing. See NOTE-section for configuration.",
daee9b
+        "required" : "0",
daee9b
+        "order" : 5
daee9b
+    }
daee9b
+    all_opt["msi"] = {
daee9b
+        "getopt" : "",
daee9b
+        "longopt" : "msi",
daee9b
+        "help" : "--msi                          Use Managed Service Identity instead of\n\
daee9b
+                                  username and password. If specified,\n\
daee9b
+                                  parameters tenantId, login and passwd are not\n\
daee9b
+                                  allowed.",
daee9b
+        "shortdesc" : "Determines if Managed Service Identity should be used.",
daee9b
+        "required" : "0",
daee9b
+        "order" : 6
daee9b
+    }
daee9b
+    all_opt["cloud"] = {
daee9b
+        "getopt" : ":",
daee9b
+        "longopt" : "cloud",
daee9b
+        "help" : "--cloud=[name]                 Name of the cloud you want to use. Supported\n\
daee9b
+                                  values are china, germany or usgov. Do not use\n\
daee9b
+                                  this parameter if you want to use public\n\
daee9b
+                                  Azure.",
daee9b
+        "shortdesc" : "Name of the cloud you want to use.",
daee9b
+        "required" : "0",
daee9b
+        "order" : 7
daee9b
+    }
daee9b
 
daee9b
 # Main agent method
daee9b
 def main():
daee9b
     compute_client = None
daee9b
+    network_client = None
daee9b
 
daee9b
-    device_opt = ["resourceGroup", "login", "passwd", "tenantId", "subscriptionId","port"]
daee9b
+    device_opt = ["login", "passwd", "port", "resourceGroup", "tenantId", "subscriptionId", "network-fencing", "msi", "cloud"]
daee9b
 
daee9b
     atexit.register(atexit_handler)
daee9b
 
daee9b
@@ -113,36 +209,44 @@
daee9b
 \n.P\n\
daee9b
 For instructions to setup credentials see: https://docs.microsoft.com/en-us/azure/azure-resource-manager/resource-group-create-service-principal-portal\
daee9b
 \n.P\n\
daee9b
-Username and password are application ID and authentication key from \"App registrations\"."
daee9b
+Username and password are application ID and authentication key from \"App registrations\".\
daee9b
+\n.P\n\
daee9b
+NOTE: NETWORK FENCING\n.br\n\
daee9b
+Network fencing requires an additional Subnet named \"fence-subnet\" for the Virtual Network using a Network Security Group with the following rules:\n.br\n\
daee9b
++-----------+-----+-------------------------+------+------+-----+-----+--------+\n.br\n\
daee9b
+| DIRECTION | PRI | NAME                    | PORT | PROT | SRC | DST | ACTION |\n.br\n\
daee9b
++-----------+-----+-------------------------+------+------+-----+-----+--------+\n.br\n\
daee9b
+| Inbound   | 100 | FENCE_DENY_ALL_INBOUND  | Any  | Any  | Any | Any | Deny   |\n.br\n\
daee9b
+| Outbound  | 100 | FENCE_DENY_ALL_OUTBOUND | Any  | Any  | Any | Any | Deny   |\n.br\n\
daee9b
++-----------+-----+-------------------------+------+------+-----+-----+--------+\
daee9b
+\n.P\n\
daee9b
+When using network fencing the reboot-action will cause a quick-return once the network has been fenced (instead of waiting for the off-action to succeed). It will check the status during the monitor-action, and request power-on when the shutdown operation is complete."
daee9b
     docs["vendorurl"] = "http://www.microsoft.com"
daee9b
     show_docs(options, docs)
daee9b
 
daee9b
     run_delay(options)
daee9b
 
daee9b
     try:
daee9b
-        from azure.common.credentials import ServicePrincipalCredentials
daee9b
-        from azure.mgmt.compute import ComputeManagementClient
daee9b
-
daee9b
-        tenantid = options["--tenantId"]
daee9b
-        servicePrincipal = options["--username"]
daee9b
-        spPassword = options["--password"]
daee9b
-        subscriptionId = options["--subscriptionId"]
daee9b
-        credentials = ServicePrincipalCredentials(
daee9b
-            client_id = servicePrincipal,
daee9b
-            secret = spPassword,
daee9b
-            tenant = tenantid
daee9b
-        )
daee9b
-        compute_client = ComputeManagementClient(
daee9b
-            credentials,
daee9b
-            subscriptionId
daee9b
-        )
daee9b
+        config = azure_fence.get_azure_config(options)
daee9b
+        compute_client = azure_fence.get_azure_compute_client(config)
daee9b
+        if "--network-fencing" in options:
daee9b
+            network_client = azure_fence.get_azure_network_client(config)
daee9b
     except ImportError:
daee9b
         fail_usage("Azure Resource Manager Python SDK not found or not accessible")
daee9b
     except Exception as e:
daee9b
         fail_usage("Failed: %s" % re.sub("^, ", "", str(e)))
daee9b
 
daee9b
+    if "--network-fencing" in options:
daee9b
+        # use  off-action to quickly return off once network is fenced instead of
daee9b
+        # waiting for vm state to change
daee9b
+        if options["--action"] == "reboot":
daee9b
+            options["--action"] = "off"
daee9b
+        # check for devices to unfence in monitor-action
daee9b
+        elif options["--action"] == "monitor":
daee9b
+            check_unfence([compute_client, network_client], options)
daee9b
+
daee9b
     # Operate the fencing device
daee9b
-    result = fence_action(compute_client, options, set_power_status, get_power_status, get_nodes_list)
daee9b
+    result = fence_action([compute_client, network_client], options, set_power_status, get_power_status, get_nodes_list)
daee9b
     sys.exit(result)
daee9b
 
daee9b
 if __name__ == "__main__":
daee9b
diff -uNr a/fence/agents/lib/azure_fence.py.py b/fence/agents/lib/azure_fence.py.py
daee9b
--- a/fence/agents/lib/azure_fence.py.py	1970-01-01 01:00:00.000000000 +0100
daee9b
+++ b/fence/agents/lib/azure_fence.py.py	2018-03-13 13:08:52.818391349 +0100
daee9b
@@ -0,0 +1,353 @@
daee9b
+import logging, re, time
daee9b
+from fencing import fail_usage
daee9b
+
daee9b
+FENCE_SUBNET_NAME = "fence-subnet"
daee9b
+FENCE_INBOUND_RULE_NAME = "FENCE_DENY_ALL_INBOUND"
daee9b
+FENCE_INBOUND_RULE_DIRECTION = "Inbound"
daee9b
+FENCE_OUTBOUND_RULE_NAME = "FENCE_DENY_ALL_OUTBOUND"
daee9b
+FENCE_OUTBOUND_RULE_DIRECTION = "Outbound"
daee9b
+FENCE_STATE_OFF = "off"
daee9b
+FENCE_STATE_ON = "on"
daee9b
+FENCE_TAG_SUBNET_ID = "FENCE_TAG_SUBNET_ID"
daee9b
+FENCE_TAG_IP_TYPE = "FENCE_TAG_IP_TYPE"
daee9b
+FENCE_TAG_IP = "FENCE_TAG_IP"
daee9b
+IP_TYPE_DYNAMIC = "Dynamic"
daee9b
+MAX_RETRY = 10
daee9b
+RETRY_WAIT = 5
daee9b
+
daee9b
+#BEGIN_VERSION_GENERATION
daee9b
+RELEASE_VERSION = "New fence lib agent - test release on steroids"
daee9b
+REDHAT_COPYRIGHT = ""
daee9b
+BUILD_DATE = "March, 2008"
daee9b
+#END_VERSION_GENERATION
daee9b
+
daee9b
+class AzureSubResource:
daee9b
+    Type = None
daee9b
+    Name = None
daee9b
+
daee9b
+class AzureResource:
daee9b
+    Id = None
daee9b
+    SubscriptionId = None
daee9b
+    ResourceGroupName = None
daee9b
+    ResourceName = None
daee9b
+    SubResources = []
daee9b
+
daee9b
+class AzureConfiguration:
daee9b
+    RGName = None
daee9b
+    VMName = None
daee9b
+    SubscriptionId = None
daee9b
+    Cloud = None
daee9b
+    UseMSI = None
daee9b
+    Tenantid = None
daee9b
+    ApplicationId = None
daee9b
+    ApplicationKey = None
daee9b
+    Verbose = None
daee9b
+
daee9b
+def get_from_metadata(parameter):
daee9b
+    import requests
daee9b
+    try:
daee9b
+        r = requests.get('http://169.254.169.254/metadata/instance?api-version=2017-08-01', headers = {"Metadata":"true"})
daee9b
+        return str(r.json()["compute"][parameter])
daee9b
+    except:
daee9b
+        logging.warning("Not able to use metadata service. Am I running in Azure?")
daee9b
+
daee9b
+    return None
daee9b
+
daee9b
+def get_azure_resource(id):
daee9b
+    match = re.match('(/subscriptions/([^/]*)/resourceGroups/([^/]*))(/providers/([^/]*/[^/]*)/([^/]*))?((/([^/]*)/([^/]*))*)', id)
daee9b
+    if not match:
daee9b
+        fail_usage("{get_azure_resource} cannot parse resource id %s" % id)
daee9b
+
daee9b
+    logging.debug("{get_azure_resource} found %s matches for %s" % (len(match.groups()), id))
daee9b
+    iGroup = 0
daee9b
+    while iGroup < len(match.groups()):
daee9b
+        logging.debug("{get_azure_resource} group %s: %s" %(iGroup, match.group(iGroup)))
daee9b
+        iGroup += 1
daee9b
+
daee9b
+    resource = AzureResource()
daee9b
+    resource.Id = id
daee9b
+    resource.SubscriptionId = match.group(2)
daee9b
+    resource.SubResources = []
daee9b
+
daee9b
+    if len(match.groups()) > 3:
daee9b
+        resource.ResourceGroupName = match.group(3)
daee9b
+        logging.debug("{get_azure_resource} resource group %s" % resource.ResourceGroupName)
daee9b
+
daee9b
+    if len(match.groups()) > 6:
daee9b
+        resource.ResourceName = match.group(6)
daee9b
+        logging.debug("{get_azure_resource} resource name %s" % resource.ResourceName)
daee9b
+
daee9b
+    if len(match.groups()) > 7 and match.group(7):
daee9b
+        splits = match.group(7).split("/")
daee9b
+        logging.debug("{get_azure_resource} splitting subtypes '%s' (%s)" % (match.group(7), len(splits)))
daee9b
+        i = 1 # the string starts with / so the first split is empty
daee9b
+        while i < len(splits) - 1:
daee9b
+            logging.debug("{get_azure_resource} creating subresource with type %s and name %s" % (splits[i], splits[i+1]))
daee9b
+            subRes = AzureSubResource()
daee9b
+            subRes.Type = splits[i]
daee9b
+            subRes.Name = splits[i+1]
daee9b
+            resource.SubResources.append(subRes)
daee9b
+            i += 2
daee9b
+
daee9b
+    return resource
daee9b
+
daee9b
+def get_fence_subnet_for_config(ipConfig, network_client):
daee9b
+    subnetResource = get_azure_resource(ipConfig.subnet.id)
daee9b
+    logging.debug("{get_fence_subnet_for_config} testing virtual network %s in resource group %s for a fence subnet" %(subnetResource.ResourceName, subnetResource.ResourceGroupName))
daee9b
+    vnet = network_client.virtual_networks.get(subnetResource.ResourceGroupName, subnetResource.ResourceName)
daee9b
+    return get_subnet(vnet, FENCE_SUBNET_NAME)
daee9b
+
daee9b
+def get_subnet(vnet, subnetName):
daee9b
+    for avSubnet in vnet.subnets:
daee9b
+        logging.debug("{get_subnet} searching subnet %s testing subnet %s" % (subnetName, avSubnet.name))
daee9b
+        if (avSubnet.name.lower() == subnetName.lower()):
daee9b
+                logging.debug("{get_subnet} subnet found %s" % avSubnet)
daee9b
+                return avSubnet
daee9b
+
daee9b
+def test_fence_subnet(fenceSubnet, nic, network_client):
daee9b
+    logging.info("{test_fence_subnet}")
daee9b
+    testOk = True
daee9b
+    if not fenceSubnet:
daee9b
+        testOk = False
daee9b
+        logging.info("{test_fence_subnet} No fence subnet found for virtual network of network interface %s" % nic.id)
daee9b
+    else:
daee9b
+        if not fenceSubnet.network_security_group:
daee9b
+            testOk = False
daee9b
+            logging.info("{test_fence_subnet} Fence subnet %s has not network security group" % fenceSubnet.id)
daee9b
+        else:
daee9b
+            nsgResource = get_azure_resource(fenceSubnet.network_security_group.id)
daee9b
+            logging.info("{test_fence_subnet} Getting network security group %s in resource group %s" % (nsgResource.ResourceName, nsgResource.ResourceGroupName))
daee9b
+            nsg = network_client.network_security_groups.get(nsgResource.ResourceGroupName, nsgResource.ResourceName)
daee9b
+            inboundRule = get_inbound_rule_for_nsg(nsg)
daee9b
+            outboundRule = get_outbound_rule_for_nsg(nsg)
daee9b
+            if not outboundRule:
daee9b
+                testOk = False
daee9b
+                logging.info("{test_fence_subnet} Network Securiy Group %s of fence subnet %s has no outbound security rule that blocks all traffic" % (nsgResource.ResourceName, fenceSubnet.id))
daee9b
+            elif not inboundRule:
daee9b
+                testOk = False
daee9b
+                logging.info("{test_fence_subnet} Network Securiy Group %s of fence subnet %s has no inbound security rule that blocks all traffic" % (nsgResource.ResourceName, fenceSubnet.id))
daee9b
+
daee9b
+    return testOk
daee9b
+
daee9b
+def get_inbound_rule_for_nsg(nsg):
daee9b
+    return get_rule_for_nsg(nsg, FENCE_INBOUND_RULE_NAME, FENCE_INBOUND_RULE_DIRECTION)
daee9b
+
daee9b
+def get_outbound_rule_for_nsg(nsg):
daee9b
+    return get_rule_for_nsg(nsg, FENCE_OUTBOUND_RULE_NAME, FENCE_OUTBOUND_RULE_DIRECTION)
daee9b
+
daee9b
+def get_rule_for_nsg(nsg, ruleName, direction):
daee9b
+    logging.info("{get_rule_for_nsg} Looking for security rule %s with direction %s" % (ruleName, direction))
daee9b
+    if not nsg:
daee9b
+        logging.info("{get_rule_for_nsg} Network security group not set")
daee9b
+        return None
daee9b
+
daee9b
+    for rule in nsg.security_rules:
daee9b
+        logging.info("{get_rule_for_nsg} Testing a %s securiy rule %s" % (rule.direction, rule.name))
daee9b
+        if (rule.access == "Deny") and (rule.direction == direction)  \
daee9b
+                and (rule.source_port_range == "*") and (rule.destination_port_range == "*") \
daee9b
+                and (rule.protocol == "*") and (rule.destination_address_prefix == "*") \
daee9b
+                and (rule.source_address_prefix == "*") and (rule.provisioning_state == "Succeeded") \
daee9b
+                and (rule.priority == 100) and (rule.name == ruleName):
daee9b
+            logging.info("{get_rule_for_nsg} %s rule found" % direction)
daee9b
+            return rule
daee9b
+
daee9b
+    return None
daee9b
+
daee9b
+def get_network_state(compute_client, network_client, rgName, vmName):
daee9b
+    result = FENCE_STATE_ON
daee9b
+
daee9b
+    try:
daee9b
+        vm = compute_client.virtual_machines.get(rgName, vmName, "instanceView")
daee9b
+
daee9b
+        allNICOK = True
daee9b
+        for nicRef in vm.network_profile.network_interfaces:
daee9b
+            nicresource = get_azure_resource(nicRef.id)
daee9b
+            nic = network_client.network_interfaces.get(nicresource.ResourceGroupName, nicresource.ResourceName)
daee9b
+            for ipConfig in nic.ip_configurations:
daee9b
+                logging.info("{get_network_state} Testing ip configuration %s" % ipConfig.name)
daee9b
+                fenceSubnet = get_fence_subnet_for_config(ipConfig, network_client)
daee9b
+                testOk = test_fence_subnet(fenceSubnet, nic, network_client)
daee9b
+                if not testOk:
daee9b
+                    allNICOK = False
daee9b
+                elif fenceSubnet.id.lower() != ipConfig.subnet.id.lower():
daee9b
+                    logging.info("{get_network_state} IP configuration %s is not in fence subnet (ip subnet: %s, fence subnet: %s)" % (ipConfig.name, ipConfig.subnet.id.lower(), fenceSubnet.id.lower()))
daee9b
+                    allNICOK = False
daee9b
+        if allNICOK:
daee9b
+            logging.info("{get_network_state} All IP configurations of all network interfaces are in the fence subnet. Declaring VM as off")
daee9b
+            result = FENCE_STATE_OFF
daee9b
+    except Exception as e:
daee9b
+        fail_usage("{get_network_state} Failed: %s" % e)
daee9b
+
daee9b
+    return result
daee9b
+
daee9b
+def set_network_state(compute_client, network_client, rgName, vmName, operation):
daee9b
+    import msrestazure.azure_exceptions
daee9b
+    logging.info("{set_network_state} Setting state %s for  %s in resource group %s" % (operation, vmName, rgName))
daee9b
+
daee9b
+    vm = compute_client.virtual_machines.get(rgName, vmName, "instanceView")
daee9b
+
daee9b
+    operations = []
daee9b
+    for nicRef in vm.network_profile.network_interfaces:
daee9b
+        for attempt in range(0, MAX_RETRY):
daee9b
+            try:
daee9b
+                nicresource = get_azure_resource(nicRef.id)
daee9b
+                nic = network_client.network_interfaces.get(nicresource.ResourceGroupName, nicresource.ResourceName)
daee9b
+
daee9b
+                if not nic.tags and operation == "block":
daee9b
+                    nic.tags = {}
daee9b
+
daee9b
+                logging.info("{set_network_state} Searching for tags required to unfence this virtual machine")
daee9b
+                for ipConfig in nic.ip_configurations:
daee9b
+                    if operation == "block":
daee9b
+                        fenceSubnet = get_fence_subnet_for_config(ipConfig, network_client)
daee9b
+                        testOk = test_fence_subnet(fenceSubnet, nic, network_client)
daee9b
+                        if testOk:
daee9b
+                            logging.info("{set_network_state} Changing subnet of ip config of nic %s" % nic.id)
daee9b
+                            nic.tags[("%s_%s" % (FENCE_TAG_SUBNET_ID, ipConfig.name))] = ipConfig.subnet.id
daee9b
+                            nic.tags[("%s_%s" % (FENCE_TAG_IP_TYPE, ipConfig.name))] = ipConfig.private_ip_allocation_method
daee9b
+                            nic.tags[("%s_%s" % (FENCE_TAG_IP, ipConfig.name))] = ipConfig.private_ip_address
daee9b
+                            ipConfig.subnet = fenceSubnet
daee9b
+                            ipConfig.private_ip_allocation_method = IP_TYPE_DYNAMIC
daee9b
+                        else:
daee9b
+                            fail_usage("{set_network_state} Network interface id %s does not have a network security group." % nic.id)
daee9b
+                    elif operation == "unblock":
daee9b
+                        if not nic.tags:
daee9b
+                            fail_usage("{set_network_state} IP configuration %s is missing the required resource tags (empty)" % ipConfig.name)
daee9b
+
daee9b
+                        subnetId = nic.tags.pop("%s_%s" % (FENCE_TAG_SUBNET_ID, ipConfig.name))
daee9b
+                        ipType = nic.tags.pop("%s_%s" % (FENCE_TAG_IP_TYPE, ipConfig.name))
daee9b
+                        ipAddress = nic.tags.pop("%s_%s" % (FENCE_TAG_IP, ipConfig.name))
daee9b
+
daee9b
+                        if (subnetId and ipType and (ipAddress or (ipType.lower() == IP_TYPE_DYNAMIC.lower()))):
daee9b
+                            logging.info("{set_network_state} tags found (subnetId: %s, ipType: %s, ipAddress: %s)" % (subnetId, ipType, ipAddress))
daee9b
+
daee9b
+                            subnetResource = get_azure_resource(subnetId)
daee9b
+                            vnet = network_client.virtual_networks.get(subnetResource.ResourceGroupName, subnetResource.ResourceName)
daee9b
+                            logging.info("{set_network_state} looking for subnet %s" % len(subnetResource.SubResources))
daee9b
+                            oldSubnet = get_subnet(vnet, subnetResource.SubResources[0].Name)
daee9b
+                            if not oldSubnet:
daee9b
+                                fail_usage("{set_network_state} subnet %s not found" % subnetId)
daee9b
+
daee9b
+                            ipConfig.subnet = oldSubnet
daee9b
+                            ipConfig.private_ip_allocation_method = ipType
daee9b
+                            if ipAddress:
daee9b
+                                ipConfig.private_ip_address = ipAddress
daee9b
+                        else:
daee9b
+                            fail_usage("{set_network_state} IP configuration %s is missing the required resource tags(subnetId: %s, ipType: %s, ipAddress: %s)" % (ipConfig.name, subnetId, ipType, ipAddress))
daee9b
+
daee9b
+                logging.info("{set_network_state} updating nic %s" % (nic.id))
daee9b
+                op = network_client.network_interfaces.create_or_update(nicresource.ResourceGroupName, nicresource.ResourceName, nic)
daee9b
+                operations.append(op)
daee9b
+                break
daee9b
+            except msrestazure.azure_exceptions.CloudError as cex:
daee9b
+                logging.error("{set_network_state} CloudError in attempt %s '%s'" % (attempt, cex))
daee9b
+                if cex.error and cex.error.error and cex.error.error.lower() == "PrivateIPAddressIsBeingCleanedUp":
daee9b
+                    logging.error("{set_network_state} PrivateIPAddressIsBeingCleanedUp")
daee9b
+                time.sleep(RETRY_WAIT)
daee9b
+
daee9b
+            except Exception as ex:
daee9b
+                logging.error("{set_network_state} Exception of type %s: %s" % (type(ex).__name__, ex))
daee9b
+                break
daee9b
+
daee9b
+def get_azure_config(options):
daee9b
+    config = AzureConfiguration()
daee9b
+
daee9b
+    config.RGName = options.get("--resourceGroup")
daee9b
+    config.VMName = options.get("--plug")
daee9b
+    config.SubscriptionId = options.get("--subscriptionId")
daee9b
+    config.Cloud = options.get("--cloud")
daee9b
+    config.UseMSI = "--msi" in options
daee9b
+    config.Tenantid = options.get("--tenantId")
daee9b
+    config.ApplicationId = options.get("--username")
daee9b
+    config.ApplicationKey = options.get("--password")
daee9b
+    config.Verbose = options.get("--verbose")
daee9b
+
daee9b
+    if not config.RGName:
daee9b
+        logging.info("resourceGroup not provided. Using metadata service")
daee9b
+        config.RGName = get_from_metadata("resourceGroupName")
daee9b
+
daee9b
+    if not config.SubscriptionId:
daee9b
+        logging.info("subscriptionId not provided. Using metadata service")
daee9b
+        config.SubscriptionId = get_from_metadata("subscriptionId")
daee9b
+
daee9b
+    return config
daee9b
+
daee9b
+def get_azure_cloud_environment(config):
daee9b
+    cloud_environment = None
daee9b
+    if config.Cloud:
daee9b
+        if (config.Cloud.lower() == "china"):
daee9b
+            from msrestazure.azure_cloud import AZURE_CHINA_CLOUD
daee9b
+            cloud_environment = AZURE_CHINA_CLOUD
daee9b
+        elif (config.Cloud.lower() == "germany"):
daee9b
+            from msrestazure.azure_cloud import AZURE_GERMAN_CLOUD
daee9b
+            cloud_environment = AZURE_GERMAN_CLOUD
daee9b
+        elif (config.Cloud.lower() == "usgov"):
daee9b
+            from msrestazure.azure_cloud import AZURE_US_GOV_CLOUD
daee9b
+            cloud_environment = AZURE_US_GOV_CLOUD
daee9b
+
daee9b
+    return cloud_environment
daee9b
+
daee9b
+def get_azure_credentials(config):
daee9b
+    credentials = None
daee9b
+    cloud_environment = get_azure_cloud_environment(config)
daee9b
+    if config.UseMSI and cloud_environment:
daee9b
+        from msrestazure.azure_active_directory import MSIAuthentication
daee9b
+        credentials = MSIAuthentication(cloud_environment=cloud_environment)
daee9b
+    elif config.UseMSI:
daee9b
+        from msrestazure.azure_active_directory import MSIAuthentication
daee9b
+        credentials = MSIAuthentication()
daee9b
+    elif cloud_environment:
daee9b
+        from azure.common.credentials import ServicePrincipalCredentials
daee9b
+        credentials = ServicePrincipalCredentials(
daee9b
+            client_id = config.ApplicationId,
daee9b
+            secret = config.ApplicationKey,
daee9b
+            tenant = config.Tenantid,
daee9b
+            cloud_environment=cloud_environment
daee9b
+        )
daee9b
+    else:
daee9b
+        from azure.common.credentials import ServicePrincipalCredentials
daee9b
+        credentials = ServicePrincipalCredentials(
daee9b
+            client_id = config.ApplicationId,
daee9b
+            secret = config.ApplicationKey,
daee9b
+            tenant = config.Tenantid
daee9b
+        )
daee9b
+
daee9b
+    return credentials
daee9b
+
daee9b
+def get_azure_compute_client(config):
daee9b
+    from azure.mgmt.compute import ComputeManagementClient
daee9b
+
daee9b
+    cloud_environment = get_azure_cloud_environment(config)
daee9b
+    credentials = get_azure_credentials(config)
daee9b
+
daee9b
+    if cloud_environment:
daee9b
+        compute_client = ComputeManagementClient(
daee9b
+            credentials,
daee9b
+            config.SubscriptionId,
daee9b
+            base_url=cloud_environment.endpoints.resource_manager
daee9b
+        )
daee9b
+    else:
daee9b
+        compute_client = ComputeManagementClient(
daee9b
+            credentials,
daee9b
+            config.SubscriptionId
daee9b
+        )
daee9b
+    return compute_client
daee9b
+
daee9b
+def get_azure_network_client(config):
daee9b
+    from azure.mgmt.network import NetworkManagementClient
daee9b
+
daee9b
+    cloud_environment = get_azure_cloud_environment(config)
daee9b
+    credentials = get_azure_credentials(config)
daee9b
+
daee9b
+    if cloud_environment:
daee9b
+        network_client = NetworkManagementClient(
daee9b
+            credentials,
daee9b
+            config.SubscriptionId,
daee9b
+            base_url=cloud_environment.endpoints.resource_manager
daee9b
+        )
daee9b
+    else:
daee9b
+        network_client = NetworkManagementClient(
daee9b
+            credentials,
daee9b
+            config.SubscriptionId
daee9b
+        )
daee9b
+    return network_client
daee9b
diff -uNr a/fence/agents/lib/Makefile.am b/fence/agents/lib/Makefile.am
daee9b
--- a/fence/agents/lib/Makefile.am	2014-08-06 09:35:08.000000000 +0200
daee9b
+++ b/fence/agents/lib/Makefile.am	2018-03-13 13:06:02.507027744 +0100
daee9b
@@ -1,12 +1,12 @@
daee9b
 MAINTAINERCLEANFILES	= Makefile.in
daee9b
 
daee9b
-TARGET			= fencing.py fencing_snmp.py
daee9b
+TARGET			= fencing.py fencing_snmp.py azure_fence.py
daee9b
 
daee9b
 if BUILD_XENAPILIB
daee9b
 TARGET			+= XenAPI.py
daee9b
 endif
daee9b
 
daee9b
-SRC			= fencing.py.py fencing_snmp.py.py XenAPI.py.py check_used_options.py
daee9b
+SRC			= fencing.py.py fencing_snmp.py.py XenAPI.py.py azure_fence.py.py check_used_options.py
daee9b
 
daee9b
 XSL			= fence2man.xsl fence2rng.xsl
daee9b