Exemple #1
0
    def _set_autoscaling(self):
        # Prepare autoscaler
        asg_manager = AXUserASGManager(self._cluster_name_id, self._region, self._aws_profile)
        asg = asg_manager.get_variable_asg() or asg_manager.get_spot_asg() or asg_manager.get_on_demand_asg()
        if not asg:
            raise AXPlatformException("Failed to get autoscaling group for cluster {}".format(self._cluster_name_id))
        asg_name = asg["AutoScalingGroupName"]

        if asg_name is not None:
            self._replacing["ASG_NAME"] = asg_name
        else:
            logger.error("Autoscaling group name not found for %s", self._cluster_name_id)
            raise AXPlatformException("Cannot find cluster autoscaling group")
 def option_to_asgs(self, option):
     """
     Returns the names of the ASGs based on the provided config option.
     """
     assert option in SpotInstanceOption.VALID_SPOT_INSTANCE_OPTIONS, \
         "{} is not a valid spot instance option".format(option)
     asg_manager = AXUserASGManager(self._cluster_name_id, self._region)
     if option == SpotInstanceOption.ALL_SPOT:
         asg_names = asg_manager.get_all_asg_names()
         return asg_names
     elif option == SpotInstanceOption.NO_SPOT:
         return []
     else:
         return [asg_manager.get_variable_asg()["AutoScalingGroupName"]]
     return
Exemple #3
0
    def modify_asg(self, min, max):
        logger.info("Modifying autoscaling group ...")

        asg_manager = AXUserASGManager(self._cluster_name_id, self._region,
                                       self._aws_profile)

        asg = asg_manager.get_variable_asg()
        if not asg:
            raise AXPlatformException(
                "Failed to get variable autoscaling group for cluster {}".
                format(self._cluster_name_id))
        asg_name = asg["AutoScalingGroupName"]
        try:
            asg_manager.set_asg_spec(name=asg_name, minsize=1, maxsize=max)
        except ClientError as ce:
            raise AXPlatformException(
                "Failed to set cluster's variable autoscaling group min/max. Error: {}"
                .format(ce))

        logger.info("Modifying cluster autoscaling group ... DONE")
Exemple #4
0
    def _generate_replacing(self):
        # Platform code are running in python 2.7, and therefore for trusted cidr list, the str() method
        # will return something like [u'54.149.149.230/32', u'73.70.250.25/32', u'104.10.248.90/32'], and
        # this 'u' prefix cannot be surpressed. With this prefix, our macro replacing would create invalid
        # yaml files, and therefore we construct string manually here
        trusted_cidr = self._cluster_config.get_trusted_cidr()
        if isinstance(trusted_cidr, list):
            trusted_cidr_str = "["
            for cidr in trusted_cidr:
                trusted_cidr_str += "\"{}\",".format(str(cidr))
            trusted_cidr_str = trusted_cidr_str[:-1]
            trusted_cidr_str += "]"
        else:
            trusted_cidr_str = "[{}]".format(trusted_cidr)

        axsys_cpu = 0
        axsys_mem = 0
        daemon_cpu = 0
        daemon_mem = 0
        for name in self._kube_objects.keys():
            cpu, mem, dcpu, dmem = self._kube_objects[name].resource_usage
            axsys_cpu += cpu
            axsys_mem += mem
            daemon_cpu += dcpu
            daemon_mem += dmem

        # kube-proxy (100m CPU and 100Mi memory. Note kube-proxy does not
        # have a memory request, but this is an approximation)
        daemon_cpu += 100
        daemon_mem += 100

        logger.info(
            "Resource Usages: axsys_cpu: %s milicores, axsys_mem: %s Mi, node_daemon_cpu: %s milicores, node_daemon_mem: %s Mi",
            axsys_cpu, axsys_mem, daemon_cpu, daemon_mem)

        axsys_node_count = int(self._cluster_config.get_asxys_node_count())
        axuser_min_count = str(
            int(self._cluster_config.get_min_node_count()) - axsys_node_count)
        axuser_max_count = str(
            int(self._cluster_config.get_max_node_count()) - axsys_node_count)
        autoscaler_scan_interval = str(
            self._cluster_config.get_autoscaler_scan_interval())

        usr_node_cpu_rsvp = float(daemon_cpu) / EC2_PARAMS[
            self._cluster_config.get_axuser_node_type()]["cpu"]
        usr_node_mem_rsvp = float(daemon_mem) / EC2_PARAMS[
            self._cluster_config.get_axuser_node_type()]["memory"]
        scale_down_util_thresh = round(
            max(usr_node_cpu_rsvp, usr_node_mem_rsvp), 3) + 0.001
        logger.info("Setting node scale down utilization threshold to %s",
                    scale_down_util_thresh)

        self._persist_node_resource_rsvp(daemon_cpu, daemon_mem)

        with open("/kubernetes/cluster/version.txt", "r") as f:
            cluster_install_version = f.read().strip()

        # Prepare autoscaler
        asg_manager = AXUserASGManager(self._cluster_name_id, self._region,
                                       self._aws_profile)
        asg = asg_manager.get_variable_asg() or asg_manager.get_spot_asg(
        ) or asg_manager.get_on_demand_asg()
        if not asg:
            raise AXPlatformException(
                "Failed to get autoscaling group for cluster {}".format(
                    self._cluster_name_id))
        asg_name = asg["AutoScalingGroupName"]

        if not asg_name:
            logger.error("Autoscaling group name not found for %s",
                         self._cluster_name_id)
            raise AXPlatformException("Cannot find cluster autoscaling group")

        # Prepare minion-manager.
        spot_instances_option = self._cluster_config.get_spot_instances_option(
        )
        minion_manager_asgs = ""
        if spot_instances_option == SpotInstanceOption.ALL_SPOT:
            for asg in asg_manager.get_all_asgs():
                minion_manager_asgs = minion_manager_asgs + asg[
                    "AutoScalingGroupName"] + " "
            minion_manager_asgs = minion_manager_asgs[:-1]
        elif spot_instances_option == SpotInstanceOption.PARTIAL_SPOT:
            minion_manager_asgs = asg_manager.get_variable_asg(
            )["AutoScalingGroupName"]

        return {
            "REGISTRY":
            self._software_info.registry,
            "REGISTRY_SECRETS":
            self._software_info.registry_secrets,
            "NAMESPACE":
            self._software_info.image_namespace,
            "VERSION":
            self._software_info.image_version,
            "AX_CLUSTER_NAME_ID":
            self._cluster_name_id,
            "AX_AWS_REGION":
            self._region,
            "AX_AWS_ACCOUNT":
            self._account,
            "AX_CUSTOMER_ID":
            AXCustomerId().get_customer_id(),
            "TRUSTED_CIDR":
            trusted_cidr_str,
            "NEW_KUBE_SALT_SHA1":
            os.getenv("NEW_KUBE_SALT_SHA1") or " ",
            "NEW_KUBE_SERVER_SHA1":
            os.getenv("NEW_KUBE_SERVER_SHA1") or " ",
            "AX_KUBE_VERSION":
            os.getenv("AX_KUBE_VERSION"),
            "AX_CLUSTER_INSTALL_VERSION":
            cluster_install_version,
            "SANDBOX_ENABLED":
            str(self._cluster_config.get_sandbox_flag()),
            "ARGO_LOG_BUCKET_NAME":
            self._cluster_config.get_support_object_store_name(),
            "ASG_MIN":
            axuser_min_count,
            "ASG_MAX":
            axuser_max_count,
            "AUTOSCALER_SCAN_INTERVAL":
            autoscaler_scan_interval,
            "SCALE_DOWN_UTIL_THRESH":
            str(scale_down_util_thresh),
            "AX_CLUSTER_META_URL_V1":
            self._bucket.get_object_url_from_key(
                key=self._cluster_config_path.cluster_metadata()),
            "ASG_NAME":
            asg_name,
            "DNS_SERVER_IP":
            os.getenv("DNS_SERVER_IP", default_kube_up_env["DNS_SERVER_IP"]),
            "AX_ENABLE_SPOT_INSTANCES":
            str(spot_instances_option != SpotInstanceOption.NO_SPOT),
            "AX_SPOT_INSTANCE_ASGS":
            minion_manager_asgs,
        }