paasta-tools 1.4.1__tar.gz → 1.4.3__tar.gz

{paasta-tools-1.4.1/paasta_tools.egg-info → paasta-tools-1.4.3}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 1.2
 Name: paasta-tools
-Version: 1.4.1
+Version: 1.4.3
 Summary: Tools for Yelps SOA infrastructure
 Home-page: UNKNOWN
 Author: Compute Infrastructure @ Yelp

{paasta-tools-1.4.1 → paasta-tools-1.4.3}/paasta_tools/__init__.py

@@ -17,4 +17,4 @@
 # setup phase, the dependencies may not exist on disk yet.
 #
 # Don't bump version manually. See `make release` docs in ./Makefile
-__version__ = "1.4.1"
+__version__ = "1.4.3"

{paasta-tools-1.4.1 → paasta-tools-1.4.3}/paasta_tools/cli/cmds/spark_run.py

@@ -358,9 +358,9 @@ def add_subparser(subparsers):
     aws_group.add_argument(
         "--aws-profile",
         help="Name of the AWS profile to load credentials from. Only used when "
-        "--aws-credentials-yaml is not specified or the service name provided by "
-        "--service does not have credentials in /etc/boto_cfg. In the latter case, "
-        "the `default` user profile will be used.",
+        "--aws-credentials-yaml is not specified and --service is either "
+        "not specified or the service does not have credentials in "
+        "/etc/boto_cfg",
         default=DEFAULT_AWS_PROFILE,
     )
 

{paasta-tools-1.4.1 → paasta-tools-1.4.3}/paasta_tools/contrib/service_shard_update.py

@@ -203,13 +203,16 @@ def main(args):
                         },
                     }
                     if args.metrics_provider is not None or args.setpoint is not None:
-                        instance_config["autoscaling"] = {}
+                        instance_config["autoscaling"] = {"metrics_providers": []}
+                        metrics_provider_config = {}
                         if args.metrics_provider is not None:
-                            instance_config["autoscaling"][
-                                "metrics_provider"
-                            ] = args.metrics_provider
+                            metrics_provider_config["type"] = args.metrics_provider
                         if args.setpoint is not None:
-                            instance_config["autoscaling"]["setpoint"] = args.setpoint
+                            metrics_provider_config["setpoint"] = args.setpoint
+                        instance_config["autoscaling"]["metrics_providers"].append(
+                            metrics_provider_config
+                        )
+
                     if args.cpus is not None:
                         instance_config["cpus"] = args.cpus
                     if args.mem is not None:

{paasta-tools-1.4.1 → paasta-tools-1.4.3}/paasta_tools/spark_tools.py

@@ -6,7 +6,6 @@ from typing import Any
 from typing import cast
 from typing import Dict
 from typing import List
-from typing import Literal
 from typing import Mapping
 from typing import Set
 
@@ -24,7 +23,6 @@ SPARK_DRIVER_POOL = "stable"
 SPARK_JOB_USER = "TRON"
 SPARK_PROMETHEUS_SHARD = "ml-compute"
 SPARK_DNS_POD_TEMPLATE = "/nail/srv/configs/spark_dns_pod_template.yaml"
-MEM_MULTIPLIER = {"k": 1024, "m": 1024**2, "g": 1024**3, "t": 1024**4}
 
 log = logging.getLogger(__name__)
 
@@ -249,22 +247,3 @@ def get_spark_driver_monitoring_labels(
         "spark.yelp.com/driver_ui_port": ui_port_str,
     }
     return labels
-
-
-def get_spark_memory_in_unit(mem: str, unit: Literal["k", "m", "g", "t"]) -> float:
-    """
-    Converts Spark memory to the desired unit.
-    mem is the same format as JVM memory strings: just number or number followed by 'k', 'm', 'g' or 't'.
-    unit can be 'k', 'm', 'g' or 't'.
-    Returns memory as a float converted to the desired unit.
-    """
-    try:
-        memory_bytes = float(mem)
-    except ValueError:
-        try:
-            memory_bytes = float(mem[:-1]) * MEM_MULTIPLIER[mem[-1]]
-        except (ValueError, IndexError):
-            print(f"Unable to parse memory value {mem}. Defaulting to 2 GB.")
-            memory_bytes = 2147483648  # default to 2 GB
-    memory_unit = memory_bytes / MEM_MULTIPLIER[unit]
-    return memory_unit

{paasta-tools-1.4.1 → paasta-tools-1.4.3}/paasta_tools/tron_tools.py

@@ -282,35 +282,6 @@ class TronActionConfig(InstanceConfig):
         self.job, self.action = decompose_instance(instance)
         # Indicate whether this config object is created for validation
         self.for_validation = for_validation
-        self.action_spark_config = None
-        if self.get_executor() == "spark":
-            # build the complete Spark configuration
-            # TODO: add conditional check for Spark specific commands spark-submit, pyspark etc ?
-            self.action_spark_config = self.build_spark_config()
-
-    def get_cpus(self) -> float:
-        # set Spark driver pod CPU if it is specified by Spark arguments
-        if (
-            self.action_spark_config
-            and "spark.driver.cores" in self.action_spark_config
-        ):
-            return float(self.action_spark_config["spark.driver.cores"])
-        # we fall back to this default if there's no spark.driver.cores config
-        return super().get_cpus()
-
-    def get_mem(self) -> float:
-        # set Spark driver pod memory if it is specified by Spark arguments
-        if (
-            self.action_spark_config
-            and "spark.driver.memory" in self.action_spark_config
-        ):
-            return int(
-                spark_tools.get_spark_memory_in_unit(
-                    self.action_spark_config["spark.driver.memory"], "m"
-                )
-            )
-        # we fall back to this default if there's no spark.driver.memory config
-        return super().get_mem()
 
     def build_spark_config(self) -> Dict[str, str]:
         system_paasta_config = load_system_paasta_config()
@@ -469,6 +440,7 @@ class TronActionConfig(InstanceConfig):
         system_paasta_config: Optional["SystemPaastaConfig"] = None,
     ) -> Dict[str, str]:
         env = super().get_env(system_paasta_config=system_paasta_config)
+
         if self.get_executor() == "spark":
             # Required by some sdks like boto3 client. Throws NoRegionError otherwise.
             # AWS_REGION takes precedence if set.
@@ -633,20 +605,6 @@ class TronActionConfig(InstanceConfig):
             error_msgs.append(
                 f"{self.get_job_name()}.{self.get_action_name()} must have a deploy_group set"
             )
-        # We are not allowing users to specify `cpus` and `mem` configuration if the action is a Spark job
-        # with driver running on k8s (executor: spark), because we derive these values from `spark.driver.cores`
-        # and `spark.driver.memory` in order to avoid confusion.
-        if self.get_executor() == "spark":
-            if "cpus" in self.config_dict:
-                error_msgs.append(
-                    f"{self.get_job_name()}.{self.get_action_name()} is a Spark job. `cpus` config is not allowed. "
-                    f"Please specify the driver cores using `spark.driver.cores`."
-                )
-            if "mem" in self.config_dict:
-                error_msgs.append(
-                    f"{self.get_job_name()}.{self.get_action_name()} is a Spark job. `mem` config is not allowed. "
-                    f"Please specify the driver memory using `spark.driver.memory`."
-                )
         return error_msgs
 
     def get_pool(self) -> str:
@@ -1017,26 +975,21 @@ def format_tron_action_dict(action_config: TronActionConfig):
         if executor == "spark":
             is_mrjob = action_config.config_dict.get("mrjob", False)
             system_paasta_config = load_system_paasta_config()
-            # inject additional Spark configs in case of Spark commands
+            # inject spark configs to the original spark-submit command
+            spark_config = action_config.build_spark_config()
             result["command"] = spark_tools.build_spark_command(
                 result["command"],
-                action_config.action_spark_config,
+                spark_config,
                 is_mrjob,
                 action_config.config_dict.get(
                     "max_runtime", spark_tools.DEFAULT_SPARK_RUNTIME_TIMEOUT
                 ),
             )
-            # point to the KUBECONFIG needed by Spark driver
             result["env"]["KUBECONFIG"] = system_paasta_config.get_spark_kubeconfig()
-
             # spark, unlike normal batches, needs to expose several ports for things like the spark
             # ui and for executor->driver communication
             result["ports"] = list(
-                set(
-                    spark_tools.get_spark_ports_from_config(
-                        action_config.action_spark_config
-                    )
-                )
+                set(spark_tools.get_spark_ports_from_config(spark_config))
             )
             # mount KUBECONFIG file for Spark drivers to communicate with EKS cluster
             extra_volumes.append(
@@ -1050,12 +1003,10 @@ def format_tron_action_dict(action_config: TronActionConfig):
             )
             # Add pod annotations and labels for Spark monitoring metrics
             monitoring_annotations = (
-                spark_tools.get_spark_driver_monitoring_annotations(
-                    action_config.action_spark_config
-                )
+                spark_tools.get_spark_driver_monitoring_annotations(spark_config)
             )
             monitoring_labels = spark_tools.get_spark_driver_monitoring_labels(
-                action_config.action_spark_config
+                spark_config
             )
             result["annotations"].update(monitoring_annotations)
             result["labels"].update(monitoring_labels)

{paasta-tools-1.4.1 → paasta-tools-1.4.3/paasta_tools.egg-info}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 1.2
 Name: paasta-tools
-Version: 1.4.1
+Version: 1.4.3
 Summary: Tools for Yelps SOA infrastructure
 Home-page: UNKNOWN
 Author: Compute Infrastructure @ Yelp

{paasta-tools-1.4.1 → paasta-tools-1.4.3}/paasta_tools.egg-info/requires.txt

@@ -46,7 +46,7 @@ requests-cache>=0.4.10
 retry
 ruamel.yaml
 sensu-plugin
-service-configuration-lib>=2.18.18
+service-configuration-lib>=2.18.17
 signalfx
 slackclient>=1.2.1
 sticht>=1.1.0

{paasta-tools-1.4.1 → paasta-tools-1.4.3}/requirements-minimal.txt

@@ -53,7 +53,7 @@ requests-cache >= 0.4.10
 retry
 ruamel.yaml
 sensu-plugin
-service-configuration-lib >= 2.18.18
+service-configuration-lib >= 2.18.17
 signalfx
 slackclient >= 1.2.1
 sticht >= 1.1.0