diff options
author | Bin Yang <bin.yang@windriver.com> | 2019-04-02 10:32:20 +0000 |
---|---|---|
committer | Bin Yang <bin.yang@windriver.com> | 2019-04-02 10:32:20 +0000 |
commit | 56b4c5cefb65f1cb76445a621087a3f4ab2fc3e9 (patch) | |
tree | bc1aa34c136b8fdba37468e7ea462c5f9629ca85 /share/starlingx_base/resource/infra_workload.py | |
parent | feec0bcd53f24b0383f9f2db3a7837fa08525947 (diff) |
Integrate heatbridge logic to infra_workload API
Offload the heatbridge to thread
Change-Id: I66c32e7ee30a40542e6ba75087d7d88caaf0c640
Issue-ID: MULTICLOUD-561
Signed-off-by: Bin Yang <bin.yang@windriver.com>
Diffstat (limited to 'share/starlingx_base/resource/infra_workload.py')
-rw-r--r-- | share/starlingx_base/resource/infra_workload.py | 619 |
1 files changed, 619 insertions, 0 deletions
diff --git a/share/starlingx_base/resource/infra_workload.py b/share/starlingx_base/resource/infra_workload.py new file mode 100644 index 00000000..4da0a0d0 --- /dev/null +++ b/share/starlingx_base/resource/infra_workload.py @@ -0,0 +1,619 @@ +# Copyright (c) 2017-2018 Wind River Systems, Inc. +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +import logging +import json +import traceback +import uuid + +from django.conf import settings +from common.exceptions import VimDriverNewtonException +# from newton_base.util import VimDriverUtils + +from keystoneauth1.exceptions import HttpError +from rest_framework import status +from rest_framework.response import Response +from rest_framework.views import APIView +from common.msapi import extsys +from common.msapi.helper import Helper as helper + +from common.utils import restcall +from newton_base.registration import registration as newton_registration +from newton_base.resource import infra_workload as newton_infra_workload +from newton_base.util import VimDriverUtils + +logger = logging.getLogger(__name__) + +# global var: Audition thread +# the id is the workloadid, which implies post to workloadid1 followed by delete workloadid1 +# will replace the previous backlog item +gInfraWorkloadThread = helper.MultiCloudThreadHelper() + +class InfraWorkload(newton_infra_workload.InfraWorkload): + def __init__(self): + self._logger = logger + + def post(self, request, vimid="", workloadid=""): + self._logger.info("vimid: %s, stackid:%s" % (vimid, workloadid)) + self._logger.info("data: %s" % (request.data)) + self._logger.debug("META: %s" % request.META) + + try: + resp_template = {} + os_status = status.HTTP_500_INTERNAL_SERVER_ERROR + + if workloadid == "": + return super(InfraWorkload, self).post(request, vimid) + else: + # a post to heatbridge + worker_self = InfraWorkloadHelper( + settings.MULTICLOUD_API_V1_PREFIX, + settings.AAI_BASE_URL + ) + backlog_item = { + "id": workloadid, + "worker": worker_self.workload_update(), + "payload": (worker_self, vimid, workloadid, request.data), + "repeat": 0, # one time job + # format of status: retcode:0 is ok, otherwise error code from http status, Status ENUM, Message + "status": (0, "AAI_UPDATE_IN_PROGRESS", "Pending to schedule") + } + gInfraWorkloadThread.add(backlog_item) + if 0 == gInfraWorkloadThread.state(): + gInfraWorkloadThread.start() + + # now query the progress + backlog_item = gInfraWorkloadThread.get(workloadid) + if not backlog_item: + # backlog item not found + return Response( + data={ + 'workload_status': "AAI_UPDATE_FAIL", + "message": "AAI update failed" + }, + status=status.HTTP_500_INTERNAL_SERVER_ERROR + ) + else: + progress = backlog_item.get("status", "Status not found") + progress_code = progress[0] + progress_status = progress[1] + progress_msg = progress[2] + return Response(data={'workload_status': progress_status, "message": progress_msg}, + status=status.HTTP_201_CREATED + if progress_code == 0 else progress_code + ) + except Exception as e: + self._logger.error(traceback.format_exc()) + return Response(data={'error': str(e)}, + status=status.HTTP_500_INTERNAL_SERVER_ERROR) + + def get(self, request, vimid="", workloadid=""): + self._logger.info("vimid, workload id: %s, %s" % (vimid, workloadid)) + self._logger.debug("META: %s" % request.META) + + try: + + if workloadid == "": + raise VimDriverNewtonException( + message="workload_id is not specified", + content="workload_id must be specified to delete the workload", + status_code = status.HTTP_400_BAD_REQUEST) + + # now query the progress + backlog_item = gInfraWorkloadThread.get(workloadid) + if not backlog_item: + # backlog item not found + return Response( + data={ + 'workload_status': "AAI_UPDATE_FAIL", + "message": "AAI update failed" + }, + status=status.HTTP_500_INTERNAL_SERVER_ERROR + ) + else: + progress = backlog_item.get("status", "Status not found") + progress_code = progress[0] + progress_status = progress[1] + progress_msg = progress[2] + return Response(data={'workload_status': progress_status, "message": progress_msg}, + status=status.HTTP_201_CREATED + if progress_code == 0 else progress_code + ) + + except Exception as e: + self._logger.error(traceback.format_exc()) + return Response(data={'error': str(e)}, + status=status.HTTP_500_INTERNAL_SERVER_ERROR) + + def delete(self, request, vimid="", workloadid=""): + self._logger.info("vimid, workload id: %s, %s" % (vimid, workloadid)) + self._logger.debug("META: %s" % request.META) + + try: + + super(InfraWorkload, self).delete(request, vimid, workloadid) + + if workloadid == "": + raise VimDriverNewtonException( + message="workload_id is not specified", + content="workload_id must be specified to delete the workload", + status_code = status.HTTP_400_BAD_REQUEST) + + # a post to heatbridge delete + worker_self = InfraWorkloadHelper( + settings.MULTICLOUD_API_V1_PREFIX, + settings.AAI_BASE_URL + ) + backlog_item = { + "id": workloadid, + "worker": worker_self.workload_delete(), + "payload": (worker_self, vimid, workloadid, request.data), + "repeat": 0, # one time job + # format of status: retcode:0 is ok, otherwise error code from http status, Status ENUM, Message + "status": (0, "AAI_REMOVE_IN_PROGRESS", "Pending to schedule") + } + gInfraWorkloadThread.add(backlog_item) + if 0 == gInfraWorkloadThread.state(): + gInfraWorkloadThread.start() + + # now query the progress + backlog_item = gInfraWorkloadThread.get(workloadid) + if not backlog_item: + # backlog item not found + return Response( + data={ + 'workload_status': "STACK_REMOVE_FAILED", + "message": "AAI update failed" + }, + status=status.HTTP_500_INTERNAL_SERVER_ERROR + ) + else: + progress = backlog_item.get("status", "Status not found") + progress_code = progress[0] + progress_status = progress[1] + progress_msg = progress[2] + return Response(data={'workload_status': progress_status, "message": progress_msg}, + status=status.HTTP_200_OK + if progress_code == 0 else progress_code + ) + + # # assume the workload_type is heat + # stack_id = workloadid + # cloud_owner, regionid = extsys.decode_vim_id(vimid) + # # should go via multicloud proxy so that + # # the selflink is updated by multicloud + # retcode, v2_token_resp_json, os_status = \ + # helper.MultiCloudIdentityHelper( + # settings.MULTICLOUD_API_V1_PREFIX, + # cloud_owner, regionid, "/v2.0/tokens") + # + # if retcode > 0 or not v2_token_resp_json: + # logger.error("authenticate fails:%s, %s, %s" % + # (cloud_owner, regionid, v2_token_resp_json)) + # return + # # tenant_id = v2_token_resp_json["access"]["token"]["tenant"]["id"] + # # tenant_name = v2_token_resp_json["access"]["token"]["tenant"]["name"] + # + # # get stack status + # service_type = "orchestration" + # resource_uri = "/stacks?id=%s" % stack_id if stack_id else "/stacks" + # self._logger.info("retrieve stack resources, URI:%s" % resource_uri) + # retcode, content, os_status = \ + # helper.MultiCloudServiceHelper(cloud_owner, regionid, + # v2_token_resp_json, + # service_type, resource_uri, + # None, "GET") + # + # stacks = content.get('stacks', []) \ + # if retcode == 0 and content else [] + # # assume there is at most 1 stack returned + # # since it was filtered by id + # stack1 = stacks[0] if stacks else None + # stack_status = "" + # + # if stack1 and 'CREATE_COMPLETE' == stack1['stack_status']: + # # delete the stack + # resource_uri = "/stacks/%s/%s" % \ + # (stack1['stack_name'], stack1['id']) + # self._logger.info("delete stack, URI:%s" % resource_uri) + # retcode, content, os_status = \ + # helper.MultiCloudServiceHelper(cloud_owner, regionid, + # v2_token_resp_json, + # service_type, resource_uri, + # None, "DELETE") + # # if retcode == 0: + # # stack_status = "DELETE_IN_PROCESS" + # # # and update AAI inventory by heatbridge-delete + # # self.heatbridge_delete(request, vimid, stack1['id']) + # + # # stub response + # resp_template = { + # "template_type": "HEAT", + # "workload_id": stack_id, + # "workload_status": stack_status + # } + # + # if retcode > 0: + # resp_template["workload_response"] = content + # + # self._logger.info("RESP with data> result:%s" % resp_template) + # return Response(status=os_status) + # except VimDriverNewtonException as e: + # self._logger.error("Plugin exception> status:%s,error:%s" + # % (e.status_code, e.content)) + # return Response(data={'error': e.content}, status=e.status_code) + # except HttpError as e: + # self._logger.error("HttpError: status:%s, response:%s" % + # (e.http_status, e.response.json())) + # return Response(data=e.response.json(), status=e.http_status) + except Exception as e: + self._logger.error(traceback.format_exc()) + return Response(data={'error': str(e)}, + status=status.HTTP_500_INTERNAL_SERVER_ERROR) + + +class APIv1InfraWorkload(InfraWorkload): + def __init__(self): + super(APIv1InfraWorkload, self).__init__() + # self._logger = logger + + def post(self, request, cloud_owner="", cloud_region_id=""): + # self._logger.info("cloud owner, cloud region id, data: %s,%s, %s" % + # (cloud_owner, cloud_region_id, request.data)) + # self._logger.debug("META: %s" % request.META) + + vimid = extsys.encode_vim_id(cloud_owner, cloud_region_id) + return super(APIv1InfraWorkload, self).post(request, vimid) + + def get(self, request, cloud_owner="", cloud_region_id="", requri=""): + # self._logger.info("cloud owner, cloud region id, data: %s,%s, %s" % + # (cloud_owner, cloud_region_id, request.data)) + # self._logger.debug("META: %s" % request.META) + + vimid = extsys.encode_vim_id(cloud_owner, cloud_region_id) + return super(APIv1InfraWorkload, self).get(request, vimid, requri) + + def delete(self, request, cloud_owner="", cloud_region_id="", requri=""): + # self._logger.info("cloud owner, cloud region id, data: %s,%s, %s" % + # (cloud_owner, cloud_region_id, request.data)) + # self._logger.debug("META: %s" % request.META) + + vimid = extsys.encode_vim_id(cloud_owner, cloud_region_id) + return super(APIv1InfraWorkload, self).delete(request, vimid, requri) + + +class InfraWorkloadHelper(newton_registration.RegistryHelper): + + def __init__(self, multicloud_prefix, aai_base_url, vimid, workloadid=""): + self.proxy_prefix = multicloud_prefix + self.aai_base_url = aai_base_url + self._logger = logger + self.vimid = vimid + self.workloadid = workloadid + super(InfraWorkloadHelper, self).__init__() + + def workload_create(self, vimid, workload_data): + data = workload_data + oof_directive = data.get("oof_directives", {}) + template_type = data.get("template_type", None) + template_data = data.get("template_data", {}) + resp_template = None + if template_type and "heat" == template_type.lower(): + # update heat parameters from oof_directive + parameters = template_data.get("parameters", {}) + + for directive in oof_directive.get("directives", []): + if directive["type"] == "vnfc": + for directive2 in directive.get("directives", []): + if directive2["type"] in ["flavor_directives", + "sriovNICNetwork_directives"]: + for attr in directive2.get("attributes", []): + flavor_label = attr.get("attribute_name", None) + flavor_value = attr.get("attribute_value", None) + if flavor_label in parameters: + parameters[flavor_label] = flavor_value + else: + self._logger.warn( + "There is no parameter exist: %s" % + flavor_label) + + # update parameters + template_data["parameters"] = parameters + + # reset to make sure "files" are empty + template_data["files"] = {} + + # authenticate + cloud_owner, regionid = extsys.decode_vim_id(vimid) + # should go via multicloud proxy so that + # the selflink is updated by multicloud + retcode, v2_token_resp_json, os_status = \ + helper.MultiCloudIdentityHelper( + settings.MULTICLOUD_API_V1_PREFIX, + cloud_owner, regionid, "/v2.0/tokens" + ) + + if retcode > 0 or not v2_token_resp_json: + logger.error("authenticate fails:%s,%s, %s" % + (cloud_owner, regionid, v2_token_resp_json)) + return ( + retcode, + "authenticate fails:%s,%s, %s" % + (cloud_owner, regionid, v2_token_resp_json) + ) + + # tenant_id = v2_token_resp_json["access"]["token"]["tenant"]["id"] + + service_type = "orchestration" + resource_uri = "/stacks" + self._logger.info("retrieve stack resources, URI:%s" % resource_uri) + retcode, content, os_status = \ + helper.MultiCloudServiceHelper(cloud_owner, regionid, + v2_token_resp_json, + service_type, resource_uri, + template_data, "POST") + + stack1 = content.get('stack', None) \ + if retcode == 0 and content else None + + stackid = stack1["id"] if stack1 else "" + resp_template = { + "template_type": template_type, + "workload_id": stackid, + "template_response": content + } + self._logger.info("RESP with data> result:%s" % resp_template) + return (0, resp_template) + + def workload_update(self, vimid, stack_id, otherinfo): + ''' + update heat resource to AAI for the specified cloud region and tenant + The resources includes: vserver, vserver/l-interface, + ''' + + cloud_owner, regionid = extsys.decode_vim_id(vimid) + # should go via multicloud proxy so that the selflink is updated by multicloud + retcode, v2_token_resp_json, os_status = \ + helper.MultiCloudIdentityHelper(settings.MULTICLOUD_API_V1_PREFIX, + cloud_owner, regionid, "/v2.0/tokens") + if retcode > 0: + logger.error("authenticate fails:%s, %s, %s" % + (cloud_owner, regionid, v2_token_resp_json)) + + return (retcode, "AAI_UPDATE_FAIL", "authenticate fails:%s, %s, %s" % + (cloud_owner, regionid, v2_token_resp_json)) + tenant_id = v2_token_resp_json["access"]["token"]["tenant"]["id"] + # tenant_name = v2_token_resp_json["access"]["token"]["tenant"]["name"] + + # common prefix + aai_cloud_region = \ + "/cloud-infrastructure/cloud-regions/cloud-region/%s/%s/tenants/tenant/%s" \ + % (cloud_owner, regionid, tenant_id) + + # get stack resource + service_type = "orchestration" + resource_uri = "/stacks/%s/resources" % (stack_id) + self._logger.info("retrieve stack resources, URI:%s" % resource_uri) + retcode, content, os_status = \ + helper.MultiCloudServiceHelper(cloud_owner, regionid, + v2_token_resp_json, + service_type, resource_uri, + None, "GET") + + resources = content.get('resources', []) if retcode == 0 and content else [] + + # find and update resources + transactions = [] + for resource in resources: + if resource.get('resource_status', None) != "CREATE_COMPLETE": + # this resource is not ready yet + continue + if resource.get('resource_type', None) == 'OS::Nova::Server': + # retrieve vserver details + service_type = "compute" + resource_uri = "/servers/%s" % (resource['physical_resource_id']) + self._logger.info("retrieve vserver detail, URI:%s" % resource_uri) + retcode, content, os_status = \ + helper.MultiCloudServiceHelper(cloud_owner, regionid, + v2_token_resp_json, + service_type, resource_uri, + None, "GET") + + self._logger.debug(" resp data:%s" % content) + vserver_detail = content.get('server', None) if retcode == 0 and content else None + if vserver_detail: + # compose inventory entry for vserver + vserver_link = "" + for link in vserver_detail['links']: + if link['rel'] == 'self': + vserver_link = link['href'] + break + pass + + # note: relationship-list to flavor/image is not be update yet + # note: volumes is not updated yet + # note: relationship-list to vnf will be handled somewhere else + aai_resource = { + 'body': { + 'vserver-name': vserver_detail['name'], + 'vserver-name2': vserver_detail['name'], + "vserver-id": vserver_detail['id'], + "vserver-selflink": vserver_link, + "prov-status": vserver_detail['status'] + }, + "uri": aai_cloud_region + "/vservers/vserver/%s" % (vserver_detail['id']) + } + + try: + # then update the resource + retcode, content, status_code = \ + restcall.req_to_aai(aai_resource['uri'], + "PUT", content=aai_resource['body']) + + if retcode == 0 and content: + content = json.JSONDecoder().decode(content) + self._logger.debug("AAI update %s response: %s" % + (aai_resource['uri'], content)) + except Exception: + self._logger.error(traceback.format_exc()) + pass + + aai_resource_transactions = {"put": [aai_resource]} + transactions.append(aai_resource_transactions) + # self._logger.debug("aai_resource :%s" % aai_resource_transactions) + pass + + for resource in resources: + if resource.get('resource_status', None) != "CREATE_COMPLETE": + continue + if resource.get('resource_type', None) == 'OS::Neutron::Port': + # retrieve vport details + service_type = "network" + resource_uri = "/v2.0/ports/%s" % (resource['physical_resource_id']) + self._logger.info("retrieve vport detail, URI:%s" % resource_uri) + retcode, content, os_status = \ + helper.MultiCloudServiceHelper(cloud_owner, regionid, + v2_token_resp_json, + service_type, resource_uri, + None, "GET") + + self._logger.debug(" resp data:%s" % content) + + vport_detail = content.get('port', None) if retcode == 0 and content else None + if vport_detail: + # compose inventory entry for vport + # note: l3-interface-ipv4-address-list, + # l3-interface-ipv6-address-list are not updated yet + # note: network-name is not update yet since the detail + # coming with network-id + aai_resource = { + "body": { + "interface-name": vport_detail['name'], + "interface-id": vport_detail['id'], + "macaddr": vport_detail['mac_address'] + }, + 'uri': + aai_cloud_region + "/vservers/vserver/%s/l-interfaces/l-interface/%s" + % (vport_detail['device_id'], vport_detail['name']) + } + try: + # then update the resource + retcode, content, status_code = \ + restcall.req_to_aai(aai_resource['uri'], "PUT", + content=aai_resource['body']) + + if retcode == 0 and content: + content = json.JSONDecoder().decode(content) + self._logger.debug("AAI update %s response: %s" % + (aai_resource['uri'], content)) + except Exception: + self._logger.error(traceback.format_exc()) + pass + + aai_resource_transactions = {"put": [aai_resource]} + transactions.append(aai_resource_transactions) + # self._logger.debug("aai_resource :%s" % aai_resource_transactions) + + pass + + # aai_transactions = {"transactions": transactions} + # self._logger.debug("aai_transactions :%s" % aai_transactions) + return (retcode, "AAI_UPDATE_COMPLETE", "succeed") + + def workload_delete(self, vimid, stack_id, otherinfo): + ''' + remove heat resource from AAI for the specified cloud region and tenant + + ''' + + # enumerate the resources + cloud_owner, regionid = extsys.decode_vim_id(vimid) + # should go via multicloud proxy so that the selflink is updated by multicloud + retcode, v2_token_resp_json, os_status = \ + helper.MultiCloudIdentityHelper(settings.MULTICLOUD_API_V1_PREFIX, + cloud_owner, regionid, "/v2.0/tokens") + if retcode > 0: + logger.error("authenticate fails:%s, %s, %s" % + (cloud_owner, regionid, v2_token_resp_json)) + return None + + tenant_id = v2_token_resp_json["access"]["token"]["tenant"]["id"] + # tenant_name = v2_token_resp_json["access"]["token"]["tenant"]["name"] + + # common prefix + aai_cloud_region = \ + "/cloud-infrastructure/cloud-regions/cloud-region/%s/%s/tenants/tenant/%s" \ + % (cloud_owner, regionid, tenant_id) + + # get stack resource + service_type = "orchestration" + resource_uri = "/stacks/%s/resources" % (stack_id) + self._logger.info("retrieve stack resources, URI:%s" % resource_uri) + retcode, content, os_status = \ + helper.MultiCloudServiceHelper(cloud_owner, regionid, + v2_token_resp_json, + service_type, resource_uri, + None, "GET") + resources = content.get('resources', []) \ + if retcode == 0 and content else [] + + vserver_list = [resource['physical_resource_id'] for resource in resources + if resource.get('resource_type', None) == 'OS::Nova::Server'] + + try: + # get list of vservers + vserver_list_url = aai_cloud_region + "/vservers?depth=all" + retcode, content, status_code = \ + restcall.req_to_aai(vserver_list_url, "GET") + if retcode > 0 or not content: + self._logger.debug("AAI get %s response: %s" % (vserver_list_url, content)) + return (retcode, "AAI_REMOVE_FAIL", "authenticate fails:%s, %s, %s" % + (cloud_owner, regionid, v2_token_resp_json)) + + content = json.JSONDecoder().decode(content) + vservers = content['vserver'] + for vserver in vservers: + if vserver['vserver-id'] not in vserver_list: + continue + + try: + # iterate vport, except will be raised if no l-interface exist + for vport in vserver['l-interfaces']['l-interface']: + # delete vport + vport_delete_url = \ + aai_cloud_region + \ + "/vservers/vserver/%s/l-interfaces/l-interface/%s?resource-version=%s" \ + % (vserver['vserver-id'], vport['interface-name'], + vport['resource-version']) + + restcall.req_to_aai(vport_delete_url, "DELETE") + except Exception: + pass + + try: + # delete vserver + vserver_delete_url = \ + aai_cloud_region + \ + "/vservers/vserver/%s?resource-version=%s" \ + % (vserver['vserver-id'], vserver['resource-version']) + + restcall.req_to_aai(vserver_delete_url, "DELETE") + except Exception: + continue + + return (retcode, "AAI_REMOVE_COMPLETE", "succeed") + except Exception: + self._logger.error(traceback.format_exc()) + return None + pass |