# Copyright 2018 Iguazio
#
# 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 getpass
import os
import traceback
import typing
import uuid
from abc import ABC, abstractmethod
from ast import literal_eval
from base64 import b64encode
from copy import deepcopy
from datetime import datetime, timedelta, timezone
from os import environ
from typing import Dict, List, Optional, Tuple, Union
import IPython
from kubernetes.client.rest import ApiException
from nuclio.build import mlrun_footer
from sqlalchemy.orm import Session
import mlrun.errors
import mlrun.utils.helpers
import mlrun.utils.regex
from mlrun.api import schemas
from mlrun.api.constants import LogSources
from mlrun.api.db.base import DBInterface
from mlrun.utils.helpers import generate_object_uri, verify_field_regex
from ..config import config
from ..datastore import store_manager
from ..db import RunDBError, get_or_set_dburl, get_run_db
from ..execution import MLClientCtx
from ..k8s_utils import get_k8s_helper
from ..kfpops import mlrun_op, write_kfpmeta
from ..lists import RunList
from ..model import (
BaseMetadata,
HyperParamOptions,
ImageBuilder,
ModelObj,
RunObject,
RunTemplate,
)
from ..secrets import SecretsStore
from ..utils import (
dict_to_json,
dict_to_yaml,
enrich_image_url,
get_in,
get_parsed_docker_registry,
get_ui_url,
is_ipython,
logger,
now_date,
update_in,
)
from .constants import PodPhases, RunStates
from .funcdoc import update_function_entry_points
from .generators import get_generator
from .utils import RunError, calc_hash, results_to_iter
run_modes = ["pass"]
spec_fields = [
"command",
"args",
"image",
"mode",
"build",
"entry_points",
"description",
"workdir",
"default_handler",
"pythonpath",
"disable_auto_mount",
"allow_empty_resources",
]
class FunctionStatus(ModelObj):
def __init__(self, state=None, build_pod=None):
self.state = state
self.build_pod = build_pod
class FunctionSpec(ModelObj):
_dict_fields = spec_fields
def __init__(
self,
command=None,
args=None,
image=None,
mode=None,
build=None,
entry_points=None,
description=None,
workdir=None,
default_handler=None,
pythonpath=None,
disable_auto_mount=False,
):
self.command = command or ""
self.image = image or ""
self.mode = mode
self.args = args or []
self.rundb = None
self.description = description or ""
self.workdir = workdir
self.pythonpath = pythonpath
self._build = None
self.build = build
self.default_handler = default_handler
# TODO: type verification (FunctionEntrypoint dict)
self.entry_points = entry_points or {}
self.disable_auto_mount = disable_auto_mount
self.allow_empty_resources = None
@property
def build(self) -> ImageBuilder:
return self._build
@build.setter
def build(self, build):
self._build = self._verify_dict(build, "build", ImageBuilder)
def enrich_function_preemption_spec(self):
pass
def validate_service_account(self, allowed_service_accounts):
pass
[docs]class BaseRuntime(ModelObj):
kind = "base"
_is_nested = False
_is_remote = False
_dict_fields = ["kind", "metadata", "spec", "status", "verbose"]
def __init__(self, metadata=None, spec=None):
self._metadata = None
self.metadata = metadata
self.kfp = None
self._spec = None
self.spec = spec
self._db_conn = None
self._secrets = None
self._k8s = None
self._is_built = False
self.is_child = False
self._status = None
self.status = None
self._is_api_server = False
self.verbose = False
[docs] def set_db_connection(self, conn, is_api=False):
if not self._db_conn:
self._db_conn = conn
self._is_api_server = is_api
@property
def metadata(self) -> BaseMetadata:
return self._metadata
@metadata.setter
def metadata(self, metadata):
self._metadata = self._verify_dict(metadata, "metadata", BaseMetadata)
@property
def spec(self) -> FunctionSpec:
return self._spec
@spec.setter
def spec(self, spec):
self._spec = self._verify_dict(spec, "spec", FunctionSpec)
@property
def status(self) -> FunctionStatus:
return self._status
@status.setter
def status(self, status):
self._status = self._verify_dict(status, "status", FunctionStatus)
def _get_k8s(self):
return get_k8s_helper()
[docs] def set_label(self, key, value):
self.metadata.labels[key] = str(value)
return self
@property
def uri(self):
return self._function_uri()
[docs] def is_deployed(self):
return True
def _is_remote_api(self):
db = self._get_db()
if db and db.kind == "http":
return True
return False
def _use_remote_api(self):
if (
self._is_remote
and not self._is_api_server
and self._get_db()
and self._get_db().kind == "http"
):
return True
return False
def _enrich_on_client_side(self):
self.try_auto_mount_based_on_config()
self.fill_credentials()
def _enrich_on_server_side(self):
pass
def _enrich_on_server_and_client_sides(self):
"""
enrich function also in client side and also on server side
"""
pass
def _enrich_function(self):
"""
enriches the function based on the flow state we run in (sdk or server)
"""
if self._use_remote_api():
self._enrich_on_client_side()
else:
self._enrich_on_server_side()
self._enrich_on_server_and_client_sides()
def _function_uri(self, tag=None, hash_key=None):
return generate_object_uri(
self.metadata.project,
self.metadata.name,
tag=tag or self.metadata.tag,
hash_key=hash_key,
)
def _ensure_run_db(self):
self.spec.rundb = self.spec.rundb or get_or_set_dburl()
def _get_db(self):
self._ensure_run_db()
if not self._db_conn:
if self.spec.rundb:
self._db_conn = get_run_db(self.spec.rundb, secrets=self._secrets)
return self._db_conn
# This function is different than the auto_mount function, as it mounts to runtimes based on the configuration.
# That's why it's named differently.
[docs] def try_auto_mount_based_on_config(self):
pass
[docs] def validate_and_enrich_service_account(
self, allowed_service_account, default_service_account
):
pass
[docs] def fill_credentials(self):
auth_session_env_var = (
mlrun.runtimes.constants.FunctionEnvironmentVariables.auth_session
)
if auth_session_env_var in os.environ or "V3IO_ACCESS_KEY" in os.environ:
self.metadata.credentials.access_key = os.environ.get(
auth_session_env_var
) or os.environ.get("V3IO_ACCESS_KEY")
[docs] def run(
self,
runspec: RunObject = None,
handler=None,
name: str = "",
project: str = "",
params: dict = None,
inputs: Dict[str, str] = None,
out_path: str = "",
workdir: str = "",
artifact_path: str = "",
watch: bool = True,
schedule: Union[str, schemas.ScheduleCronTrigger] = None,
hyperparams: Dict[str, list] = None,
hyper_param_options: HyperParamOptions = None,
verbose=None,
scrape_metrics: bool = None,
local=False,
local_code_path=None,
auto_build=None,
) -> RunObject:
"""Run a local or remote task.
:param runspec: run template object or dict (see RunTemplate)
:param handler: pointer or name of a function handler
:param name: execution name
:param project: project name
:param params: input parameters (dict)
:param inputs: input objects (dict of key: path)
:param out_path: default artifact output path
:param artifact_path: default artifact output path (will replace out_path)
:param workdir: default input artifacts path
:param watch: watch/follow run log
:param schedule: ScheduleCronTrigger class instance or a standard crontab expression string
(which will be converted to the class using its `from_crontab` constructor),
see this link for help:
https://apscheduler.readthedocs.io/en/v3.6.3/modules/triggers/cron.html#module-apscheduler.triggers.cron
:param hyperparams: dict of param name and list of values to be enumerated e.g. {"p1": [1,2,3]}
the default strategy is grid search, can specify strategy (grid, list, random)
and other options in the hyper_param_options parameter
:param hyper_param_options: dict or :py:class:`~mlrun.model.HyperParamOptions` struct of
hyper parameter options
:param verbose: add verbose prints/logs
:param scrape_metrics: whether to add the `mlrun/scrape-metrics` label to this run's resources
:param local: run the function locally vs on the runtime/cluster
:param local_code_path: path of the code for local runs & debug
:param auto_build: when set to True and the function require build it will be built on the first
function run, use only if you dont plan on changing the build config between runs
:return: run context object (RunObject) with run metadata, results and status
"""
mlrun.utils.helpers.verify_dict_items_type("Inputs", inputs, [str], [str])
if self.spec.mode and self.spec.mode not in run_modes:
raise ValueError(f'run mode can only be {",".join(run_modes)}')
self._enrich_function()
if local:
return self._run_local(
runspec,
schedule,
local_code_path,
project,
name,
workdir,
handler,
params,
inputs,
artifact_path,
)
run = self._create_run_object(runspec)
run = self._enrich_run(
run,
handler,
project,
name,
params,
inputs,
hyperparams,
hyper_param_options,
verbose,
scrape_metrics,
out_path,
artifact_path,
workdir,
)
if is_local(run.spec.output_path):
logger.warning(
"artifact path is not defined or is local,"
" artifacts will not be visible in the UI"
)
if self.kind not in ["", "local", "handler", "dask"]:
raise ValueError(
"absolute artifact_path must be specified"
" when running remote tasks"
)
db = self._get_db()
if not self.is_deployed():
if self.spec.build.auto_build or auto_build:
logger.info(
"Function is not deployed and auto_build flag is set, starting deploy..."
)
self.deploy(skip_deployed=True, show_on_failure=True)
else:
raise RunError(
"function image is not built/ready, use .deploy() method first"
)
if self.verbose:
logger.info(f"runspec:\n{run.to_yaml()}")
if "V3IO_USERNAME" in environ and "v3io_user" not in run.metadata.labels:
run.metadata.labels["v3io_user"] = environ.get("V3IO_USERNAME")
if not self.is_child:
self._store_function(run, run.metadata, db)
# execute the job remotely (to a k8s cluster via the API service)
if self._use_remote_api():
return self._submit_job(run, schedule, db, watch)
elif self._is_remote and not self._is_api_server and not self.kfp:
logger.warning(
"warning!, Api url not set, " "trying to exec remote runtime locally"
)
execution = MLClientCtx.from_dict(
run.to_dict(), db, autocommit=False, is_api=self._is_api_server
)
self._pre_run(run, execution) # hook for runtime specific prep
# create task generator (for child runs) from spec
task_generator = None
if not self._is_nested:
task_generator = get_generator(run.spec, execution)
last_err = None
if task_generator:
# multiple runs (based on hyper params or params file)
runner = self._run_many
if hasattr(self, "_parallel_run_many") and task_generator.use_parallel():
runner = self._parallel_run_many
results = runner(task_generator, execution, run)
results_to_iter(results, run, execution)
result = execution.to_dict()
else:
# single run
try:
resp = self._run(run, execution)
if watch and self.kind not in ["", "handler", "local"]:
state = run.logs(True, self._get_db())
if state != "succeeded":
logger.warning(f"run ended with state {state}")
result = self._update_run_state(resp, task=run)
except RunError as err:
last_err = err
result = self._update_run_state(task=run, err=err)
self._post_run(result, execution) # hook for runtime specific cleanup
return self._wrap_run_result(result, run, schedule=schedule, err=last_err)
def _wrap_run_result(
self, result: dict, runspec: RunObject, schedule=None, err=None
):
# if the purpose was to schedule (and not to run) nothing to wrap
if schedule:
return
if result and self.kfp and err is None:
write_kfpmeta(result)
# show ipython/jupyter result table widget
results_tbl = RunList()
if result:
results_tbl.append(result)
else:
logger.info("no returned result (job may still be in progress)")
results_tbl.append(runspec.to_dict())
uid = runspec.metadata.uid
project = runspec.metadata.project
if is_ipython and config.ipython_widget:
results_tbl.show()
print()
ui_url = get_ui_url(project, uid)
if ui_url:
ui_url = f' or <a href="{ui_url}" target="_blank">click here</a> to open in UI'
IPython.display.display(
IPython.display.HTML(
f"<b> > to track results use the .show() or .logs() methods {ui_url}</b>"
)
)
elif not self.is_child:
ui_url = get_ui_url(project, uid)
ui_url = f"\nor click {ui_url} for UI" if ui_url else ""
project_flag = f"-p {project}" if project else ""
print(
f"to track results use the CLI:\n"
f"info: mlrun get run {uid} {project_flag}\nlogs: mlrun logs {uid} {project_flag}{ui_url}"
)
if result:
run = RunObject.from_dict(result)
logger.info(f"run executed, status={run.status.state}")
if run.status.state == "error":
if self._is_remote and not self.is_child:
print(f"runtime error: {run.status.error}")
raise RunError(run.status.error)
return run
return None
def _get_db_run(self, task: RunObject = None):
if self._get_db() and task:
project = task.metadata.project
uid = task.metadata.uid
iter = task.metadata.iteration
try:
return self._get_db().read_run(uid, project, iter=iter)
except RunDBError:
return None
if task:
return task.to_dict()
def _generate_runtime_env(self, runobj: RunObject):
runtime_env = {
"MLRUN_EXEC_CONFIG": runobj.to_json(),
"MLRUN_DEFAULT_PROJECT": runobj.metadata.project
or self.metadata.project
or config.default_project,
}
if runobj.spec.verbose:
runtime_env["MLRUN_LOG_LEVEL"] = "DEBUG"
if config.httpdb.api_url:
runtime_env["MLRUN_DBPATH"] = config.httpdb.api_url
if self.metadata.namespace or config.namespace:
runtime_env["MLRUN_NAMESPACE"] = self.metadata.namespace or config.namespace
return runtime_env
def _run_local(
self,
runspec,
schedule,
local_code_path,
project,
name,
workdir,
handler,
params,
inputs,
artifact_path,
):
if schedule is not None:
raise mlrun.errors.MLRunInvalidArgumentError(
"local and schedule cannot be used together"
)
# allow local run simulation with a flip of a flag
command = self
if local_code_path:
project = project or self.metadata.project
name = name or self.metadata.name
command = local_code_path
return mlrun.run_local(
runspec,
command,
name,
self.spec.args,
workdir=workdir,
project=project,
handler=handler,
params=params,
inputs=inputs,
artifact_path=artifact_path,
mode=self.spec.mode,
allow_empty_resources=self.spec.allow_empty_resources,
)
def _create_run_object(self, runspec):
if runspec:
runspec = deepcopy(runspec)
if isinstance(runspec, str):
runspec = literal_eval(runspec)
if not isinstance(runspec, (dict, RunTemplate, RunObject)):
raise ValueError(
"task/runspec is not a valid task object," f" type={type(runspec)}"
)
if isinstance(runspec, RunTemplate):
runspec = RunObject.from_template(runspec)
if isinstance(runspec, dict) or runspec is None:
runspec = RunObject.from_dict(runspec)
return runspec
def _enrich_run(
self,
runspec,
handler,
project,
name,
params,
inputs,
hyperparams,
hyper_param_options,
verbose,
scrape_metrics,
out_path,
artifact_path,
workdir,
):
runspec.spec.handler = (
handler or runspec.spec.handler or self.spec.default_handler or ""
)
if runspec.spec.handler and self.kind not in ["handler", "dask"]:
runspec.spec.handler = runspec.spec.handler_name
def_name = self.metadata.name
if runspec.spec.handler_name:
short_name = runspec.spec.handler_name
for separator in ["#", "::", "."]:
# drop paths, module or class name from short name
if separator in short_name:
short_name = short_name.split(separator)[-1]
def_name += "-" + short_name
runspec.metadata.name = name or runspec.metadata.name or def_name
verify_field_regex(
"run.metadata.name", runspec.metadata.name, mlrun.utils.regex.run_name
)
runspec.metadata.project = (
project
or runspec.metadata.project
or self.metadata.project
or config.default_project
)
runspec.spec.parameters = params or runspec.spec.parameters
runspec.spec.inputs = inputs or runspec.spec.inputs
runspec.spec.hyperparams = hyperparams or runspec.spec.hyperparams
runspec.spec.hyper_param_options = (
hyper_param_options or runspec.spec.hyper_param_options
)
runspec.spec.verbose = verbose or runspec.spec.verbose
if scrape_metrics is None:
if runspec.spec.scrape_metrics is None:
scrape_metrics = config.scrape_metrics
else:
scrape_metrics = runspec.spec.scrape_metrics
runspec.spec.scrape_metrics = scrape_metrics
runspec.spec.output_path = out_path or artifact_path or runspec.spec.output_path
runspec.spec.input_path = (
workdir or runspec.spec.input_path or self.spec.workdir
)
if self.spec.allow_empty_resources:
runspec.spec.allow_empty_resources = self.spec.allow_empty_resources
spec = runspec.spec
if spec.secret_sources:
self._secrets = SecretsStore.from_list(spec.secret_sources)
# update run metadata (uid, labels) and store in DB
meta = runspec.metadata
meta.uid = meta.uid or uuid.uuid4().hex
runspec.spec.output_path = runspec.spec.output_path or config.artifact_path
if runspec.spec.output_path:
runspec.spec.output_path = runspec.spec.output_path.replace(
"{{run.uid}}", meta.uid
)
runspec.spec.output_path = mlrun.utils.helpers.fill_artifact_path_template(
runspec.spec.output_path, runspec.metadata.project
)
return runspec
def _submit_job(self, runspec, schedule, db, watch):
if self._secrets:
runspec.spec.secret_sources = self._secrets.to_serial()
try:
resp = db.submit_job(runspec, schedule=schedule)
if schedule:
logger.info(f"task scheduled, {resp}")
return
except Exception as err:
logger.error(f"got remote run err, {err}")
result = None
# if we got a schedule no reason to do post_run stuff (it purposed to update the run status with error,
# but there's no run in case of schedule)
if not schedule:
result = self._update_run_state(task=runspec, err=err)
return self._wrap_run_result(result, runspec, schedule=schedule, err=err)
if resp:
txt = get_in(resp, "status.status_text")
if txt:
logger.info(txt)
if watch or self.kfp:
runspec.logs(True, self._get_db())
resp = self._get_db_run(runspec)
return self._wrap_run_result(resp, runspec, schedule=schedule)
def _store_function(self, runspec, meta, db):
db_str = "self" if self._is_api_server else self.spec.rundb
logger.info(f"starting run {meta.name} uid={meta.uid} DB={db_str}")
meta.labels["kind"] = self.kind
if "owner" not in meta.labels:
meta.labels["owner"] = environ.get("V3IO_USERNAME") or getpass.getuser()
if runspec.spec.output_path:
runspec.spec.output_path = runspec.spec.output_path.replace(
"{{run.user}}", meta.labels["owner"]
)
if db and self.kind != "handler":
struct = self.to_dict()
hash_key = db.store_function(
struct, self.metadata.name, self.metadata.project, versioned=True
)
runspec.spec.function = self._function_uri(hash_key=hash_key)
def _get_cmd_args(self, runobj: RunObject):
extra_env = self._generate_runtime_env(runobj)
if self.spec.pythonpath:
extra_env["PYTHONPATH"] = self.spec.pythonpath
args = []
command = self.spec.command
code = (
self.spec.build.functionSourceCode if hasattr(self.spec, "build") else None
)
if runobj.spec.handler and self.spec.mode == "pass":
raise ValueError('cannot use "pass" mode with handler')
if code:
extra_env["MLRUN_EXEC_CODE"] = code
load_archive = self.spec.build.load_source_on_run and self.spec.build.source
need_mlrun = code or load_archive or self.spec.mode != "pass"
if need_mlrun:
args = ["run", "--name", runobj.metadata.name, "--from-env"]
if runobj.spec.handler:
args += ["--handler", runobj.spec.handler]
if self.spec.mode:
args += ["--mode", self.spec.mode]
if self.spec.build.origin_filename:
args += ["--origin-file", self.spec.build.origin_filename]
if load_archive:
if code:
raise ValueError("cannot specify both code and source archive")
args += ["--source", self.spec.build.source]
if self.spec.workdir:
# set the absolute/relative path to the cloned code
args += ["--workdir", self.spec.workdir]
if command:
args += [command]
command = "mlrun"
if self.spec.args:
args = args + self.spec.args
else:
command = command.format(**runobj.spec.parameters)
if self.spec.args:
args = [arg.format(**runobj.spec.parameters) for arg in self.spec.args]
extra_env = [{"name": k, "value": v} for k, v in extra_env.items()]
return command, args, extra_env
def _pre_run(self, runspec: RunObject, execution):
pass
def _post_run(self, results, execution):
pass
def _run(self, runobj: RunObject, execution) -> dict:
pass
def _run_many(self, generator, execution, runobj: RunObject) -> RunList:
results = RunList()
num_errors = 0
tasks = generator.generate(runobj)
for task in tasks:
try:
self.store_run(task)
resp = self._run(task, execution)
resp = self._update_run_state(resp, task=task)
run_results = resp["status"].get("results", {})
if generator.eval_stop_condition(run_results):
logger.info(
f"reached early stop condition ({generator.options.stop_condition}), stopping iterations!"
)
results.append(resp)
break
except RunError as err:
task.status.state = "error"
task.status.error = str(err)
resp = self._update_run_state(task=task, err=err)
num_errors += 1
if num_errors > generator.max_errors:
logger.error("too many errors, stopping iterations!")
results.append(resp)
break
results.append(resp)
return results
[docs] def store_run(self, runobj: RunObject):
if self._get_db() and runobj:
project = runobj.metadata.project
uid = runobj.metadata.uid
iter = runobj.metadata.iteration
self._get_db().store_run(runobj.to_dict(), uid, project, iter=iter)
def _store_run_dict(self, rundict: dict):
if self._get_db() and rundict:
project = get_in(rundict, "metadata.project", "")
uid = get_in(rundict, "metadata.uid")
iter = get_in(rundict, "metadata.iteration", 0)
self._get_db().store_run(rundict, uid, project, iter=iter)
def _update_run_state(
self,
resp: dict = None,
task: RunObject = None,
err=None,
) -> dict:
"""update the task state in the DB"""
was_none = False
if resp is None and task:
was_none = True
resp = self._get_db_run(task)
if not resp:
self.store_run(task)
return task.to_dict()
if task.status.status_text:
update_in(resp, "status.status_text", task.status.status_text)
if resp is None:
return None
if not isinstance(resp, dict):
raise ValueError(f"post_run called with type {type(resp)}")
updates = None
last_state = get_in(resp, "status.state", "")
if last_state == "error" or err:
updates = {"status.last_update": now_date().isoformat()}
updates["status.state"] = "error"
update_in(resp, "status.state", "error")
if err:
update_in(resp, "status.error", str(err))
err = get_in(resp, "status.error")
if err:
updates["status.error"] = str(err)
elif not was_none and last_state != "completed":
updates = {"status.last_update": now_date().isoformat()}
updates["status.state"] = "completed"
update_in(resp, "status.state", "completed")
if self._get_db() and updates:
project = get_in(resp, "metadata.project")
uid = get_in(resp, "metadata.uid")
iter = get_in(resp, "metadata.iteration", 0)
self._get_db().update_run(updates, uid, project, iter=iter)
return resp
def _force_handler(self, handler):
if not handler:
raise RunError(f"handler must be provided for {self.kind} runtime")
[docs] def full_image_path(self, image=None, client_version: str = None):
image = image or self.spec.image or ""
image = enrich_image_url(image, client_version)
if not image.startswith("."):
return image
registry, repository = get_parsed_docker_registry()
if registry:
if repository and repository not in image:
return f"{registry}/{repository}/{image[1:]}"
return f"{registry}/{image[1:]}"
namespace_domain = environ.get("IGZ_NAMESPACE_DOMAIN", None)
if namespace_domain is not None:
return f"docker-registry.{namespace_domain}:80/{image[1:]}"
raise RunError("local container registry is not defined")
[docs] def as_step(
self,
runspec: RunObject = None,
handler=None,
name: str = "",
project: str = "",
params: dict = None,
hyperparams=None,
selector="",
hyper_param_options: HyperParamOptions = None,
inputs: dict = None,
outputs: dict = None,
workdir: str = "",
artifact_path: str = "",
image: str = "",
labels: dict = None,
use_db=True,
verbose=None,
scrape_metrics=False,
):
"""Run a local or remote task.
:param runspec: run template object or dict (see RunTemplate)
:param handler: name of the function handler
:param name: execution name
:param project: project name
:param params: input parameters (dict)
:param hyperparams: hyper parameters
:param selector: selection criteria for hyper params
:param hyper_param_options: hyper param options (selector, early stop, strategy, ..)
see: :py:class:`~mlrun.model.HyperParamOptions`
:param inputs: input objects (dict of key: path)
:param outputs: list of outputs which can pass in the workflow
:param artifact_path: default artifact output path (replace out_path)
:param workdir: default input artifacts path
:param image: container image to use
:param labels: labels to tag the job/run with ({key:val, ..})
:param use_db: save function spec in the db (vs the workflow file)
:param verbose: add verbose prints/logs
:param scrape_metrics: whether to add the `mlrun/scrape-metrics` label to this run's resources
:return: KubeFlow containerOp
"""
# if self.spec.image and not image:
# image = self.full_image_path()
if use_db:
# if the same function is built as part of the pipeline we do not use the versioned function
# rather the latest function w the same tag so we can pick up the updated image/status
versioned = False if hasattr(self, "_build_in_pipeline") else True
url = self.save(versioned=versioned, refresh=True)
else:
url = None
if runspec is not None:
verify_field_regex(
"run.metadata.name", runspec.metadata.name, mlrun.utils.regex.run_name
)
return mlrun_op(
name,
project,
function=self,
func_url=url,
runobj=runspec,
handler=handler,
params=params,
hyperparams=hyperparams,
selector=selector,
hyper_param_options=hyper_param_options,
inputs=inputs,
outputs=outputs,
job_image=image,
labels=labels,
out_path=artifact_path,
in_path=workdir,
verbose=verbose,
scrape_metrics=scrape_metrics,
)
[docs] def with_code(self, from_file="", body=None, with_doc=True):
"""Update the function code
This function eliminates the need to build container images every time we edit the code
:param from_file: blank for current notebook, or path to .py/.ipynb file
:param body: will use the body as the function code
:param with_doc: update the document of the function parameters
:return: function object
"""
if body and from_file:
raise mlrun.errors.MLRunInvalidArgumentError(
"must provide either body or from_file argument. not both"
)
if (not body and not from_file) or (from_file and from_file.endswith(".ipynb")):
from nuclio import build_file
_, _, body = build_file(from_file, name=self.metadata.name)
else:
if from_file:
with open(from_file) as fp:
body = fp.read()
if self.kind == mlrun.runtimes.RuntimeKinds.serving:
body = body + mlrun_footer.format(
mlrun.runtimes.serving.serving_subkind
)
self.spec.build.functionSourceCode = b64encode(body.encode("utf-8")).decode(
"utf-8"
)
if with_doc:
update_function_entry_points(self, body)
return self
[docs] def with_requirements(self, requirements: Union[str, List[str]]):
"""add package requirements from file or list to build spec.
:param requirements: python requirements file path or list of packages
:return: function object
"""
if isinstance(requirements, str):
with open(requirements, "r") as fp:
requirements = fp.read().splitlines()
commands = self.spec.build.commands or []
new_command = "python -m pip install " + " ".join(requirements)
# make sure we dont append the same line twice
if new_command not in commands:
commands.append(new_command)
self.spec.build.commands = commands
self.verify_base_image()
return self
[docs] def verify_base_image(self):
build = self.spec.build
require_build = build.commands or (
build.source and not build.load_source_on_run
)
if (
self.kind not in mlrun.runtimes.RuntimeKinds.nuclio_runtimes()
and require_build
and self.spec.image
and not self.spec.build.base_image
):
# when the function require build use the image as the base_image for the build
self.spec.build.base_image = self.spec.image
self.spec.image = ""
[docs] def export(self, target="", format=".yaml", secrets=None, strip=True):
"""save function spec to a local/remote path (default to./function.yaml)
:param target: target path/url
:param format: `.yaml` (default) or `.json`
:param secrets: optional secrets dict/object for target path (e.g. s3)
:param strip: strip status data
:returns: self
"""
if self.kind == "handler":
raise ValueError(
"cannot export local handler function, use "
+ "code_to_function() to serialize your function"
)
calc_hash(self)
struct = self.to_dict(strip=strip)
if format == ".yaml":
data = dict_to_yaml(struct)
else:
data = dict_to_json(struct)
stores = store_manager.set(secrets)
target = target or "function.yaml"
datastore, subpath = stores.get_or_create_store(target)
datastore.put(subpath, data)
logger.info(f"function spec saved to path: {target}")
return self
[docs] def save(self, tag="", versioned=False, refresh=False) -> str:
db = self._get_db()
if not db:
logger.error("database connection is not configured")
return ""
if refresh and self._is_remote_api():
try:
meta = self.metadata
db_func = db.get_function(meta.name, meta.project, meta.tag)
if db_func and "status" in db_func:
self.status = db_func["status"]
if (
self.status.state
and self.status.state == "ready"
and not hasattr(self.status, "nuclio_name")
):
self.spec.image = get_in(db_func, "spec.image", self.spec.image)
except mlrun.errors.MLRunNotFoundError:
pass
tag = tag or self.metadata.tag
obj = self.to_dict()
logger.debug(f"saving function: {self.metadata.name}, tag: {tag}")
hash_key = db.store_function(
obj, self.metadata.name, self.metadata.project, tag, versioned
)
hash_key = hash_key if versioned else None
return "db://" + self._function_uri(hash_key=hash_key, tag=tag)
[docs] def to_dict(self, fields=None, exclude=None, strip=False):
struct = super().to_dict(fields, exclude=exclude)
if strip:
if "status" in struct:
del struct["status"]
return struct
[docs] def doc(self):
print("function:", self.metadata.name)
print(self.spec.description)
if self.spec.default_handler:
print("default handler:", self.spec.default_handler)
if self.spec.entry_points:
print("entry points:")
for name, entry in self.spec.entry_points.items():
print(f" {name}: {entry.get('doc', '')}")
params = entry.get("parameters")
if params:
for p in params:
line = p["name"]
if "type" in p:
line += f"({p['type']})"
line += " - " + p.get("doc", "")
if "default" in p:
line += f", default={p['default']}"
print(" " + line)
def is_local(url):
if not url:
return True
return "://" not in url and not url.startswith("/")
class BaseRuntimeHandler(ABC):
# setting here to allow tests to override
kind = "base"
wait_for_deletion_interval = 10
@staticmethod
@abstractmethod
def _get_object_label_selector(object_id: str) -> str:
"""
Should return the label selector should be used to get only resources of a specific object (with id object_id)
"""
pass
@staticmethod
@abstractmethod
def _get_possible_mlrun_class_label_values() -> List[str]:
"""
Should return the possible values of the mlrun/class label for runtime resources that are of this runtime
handler kind
"""
pass
def list_resources(
self,
project: str,
object_id: typing.Optional[str] = None,
label_selector: str = None,
group_by: Optional[mlrun.api.schemas.ListRuntimeResourcesGroupByField] = None,
) -> Union[
mlrun.api.schemas.RuntimeResources,
mlrun.api.schemas.GroupedByJobRuntimeResourcesOutput,
mlrun.api.schemas.GroupedByProjectRuntimeResourcesOutput,
]:
# We currently don't support removing runtime resources in non k8s env
if not mlrun.k8s_utils.get_k8s_helper(
silent=True
).is_running_inside_kubernetes_cluster():
return {}
k8s_helper = get_k8s_helper()
namespace = k8s_helper.resolve_namespace()
label_selector = self._resolve_label_selector(
project, object_id, label_selector
)
pods = self._list_pods(namespace, label_selector)
pod_resources = self._build_pod_resources(pods)
crd_objects = self._list_crd_objects(namespace, label_selector)
crd_resources = self._build_crd_resources(crd_objects)
response = self._build_list_resources_response(
pod_resources, crd_resources, group_by
)
response = self._enrich_list_resources_response(
response, namespace, label_selector, group_by
)
return response
def build_output_from_runtime_resources(
self,
runtime_resources_list: List[mlrun.api.schemas.RuntimeResources],
group_by: Optional[mlrun.api.schemas.ListRuntimeResourcesGroupByField] = None,
):
pod_resources = []
crd_resources = []
for runtime_resources in runtime_resources_list:
pod_resources += runtime_resources.pod_resources
crd_resources += runtime_resources.crd_resources
response = self._build_list_resources_response(
pod_resources, crd_resources, group_by
)
response = self._build_output_from_runtime_resources(
response, runtime_resources_list, group_by
)
return response
def delete_resources(
self,
db: DBInterface,
db_session: Session,
label_selector: str = None,
force: bool = False,
grace_period: int = None,
):
if grace_period is None:
grace_period = config.runtime_resources_deletion_grace_period
# We currently don't support removing runtime resources in non k8s env
if not mlrun.k8s_utils.get_k8s_helper(
silent=True
).is_running_inside_kubernetes_cluster():
return
k8s_helper = get_k8s_helper()
namespace = k8s_helper.resolve_namespace()
label_selector = self._resolve_label_selector(
"*", label_selector=label_selector
)
crd_group, crd_version, crd_plural = self._get_crd_info()
if crd_group and crd_version and crd_plural:
deleted_resources = self._delete_crd_resources(
db,
db_session,
namespace,
label_selector,
force,
grace_period,
)
else:
deleted_resources = self._delete_pod_resources(
db,
db_session,
namespace,
label_selector,
force,
grace_period,
)
self._delete_resources(
db,
db_session,
namespace,
deleted_resources,
label_selector,
force,
grace_period,
)
def delete_runtime_object_resources(
self,
db: DBInterface,
db_session: Session,
object_id: str,
label_selector: str = None,
force: bool = False,
grace_period: int = None,
):
if grace_period is None:
grace_period = config.runtime_resources_deletion_grace_period
label_selector = self._add_object_label_selector_if_needed(
object_id, label_selector
)
self.delete_resources(db, db_session, label_selector, force, grace_period)
def monitor_runs(self, db: DBInterface, db_session: Session):
k8s_helper = get_k8s_helper()
namespace = k8s_helper.resolve_namespace()
label_selector = self._get_default_label_selector()
crd_group, crd_version, crd_plural = self._get_crd_info()
runtime_resource_is_crd = False
if crd_group and crd_version and crd_plural:
runtime_resource_is_crd = True
runtime_resources = self._list_crd_objects(namespace, label_selector)
else:
runtime_resources = self._list_pods(namespace, label_selector)
project_run_uid_map = self._list_runs_for_monitoring(db, db_session)
# project -> uid -> {"name": <runtime-resource-name>}
run_runtime_resources_map = {}
for runtime_resource in runtime_resources:
project, uid, name = self._resolve_runtime_resource_run(runtime_resource)
run_runtime_resources_map.setdefault(project, {})
run_runtime_resources_map.get(project).update({uid: {"name": name}})
try:
self._monitor_runtime_resource(
db,
db_session,
project_run_uid_map,
runtime_resource,
runtime_resource_is_crd,
namespace,
project,
uid,
name,
)
except Exception as exc:
logger.warning(
"Failed monitoring runtime resource. Continuing",
runtime_resource_name=runtime_resource["metadata"]["name"],
namespace=namespace,
exc=str(exc),
traceback=traceback.format_exc(),
)
for project, runs in project_run_uid_map.items():
if runs:
for run_uid, run in runs.items():
try:
if not run:
run = db.read_run(db_session, run_uid, project)
if self.kind == run.get("metadata", {}).get("labels", {}).get(
"kind", ""
):
self._ensure_run_not_stuck_on_non_terminal_state(
db,
db_session,
project,
run_uid,
run,
run_runtime_resources_map,
)
except Exception as exc:
logger.warning(
"Failed ensuring run not stuck. Continuing",
run_uid=run_uid,
run=run,
project=project,
exc=str(exc),
traceback=traceback.format_exc(),
)
def _ensure_run_not_stuck_on_non_terminal_state(
self,
db: DBInterface,
db_session: Session,
project: str,
run_uid: str,
run: dict = None,
run_runtime_resources_map: dict = None,
):
"""
Ensuring that a run does not become trapped in a non-terminal state as a result of not finding
corresponding k8s resource.
This can occur when a node is evicted or preempted, causing the resources to be removed from the resource
listing when the final state recorded in the database is non-terminal.
This will have a significant impact on scheduled jobs, since they will not be created until the
previous run reaches a terminal state (because of concurrency limit)
"""
now = now_date()
db_run_state = run.get("status", {}).get("state")
if not db_run_state:
# we are setting the run state to a terminal state to avoid log spamming, this is mainly sanity as we are
# setting state to runs when storing new runs.
logger.info(
"Runs monitoring found a run without state, updating to a terminal state",
project=project,
uid=run_uid,
db_run_state=db_run_state,
now=now,
)
run.setdefault("status", {})["state"] = RunStates.error
run.setdefault("status", {})["last_update"] = now.isoformat()
db.store_run(db_session, run, run_uid, project)
return
if db_run_state in RunStates.non_terminal_states():
if run_runtime_resources_map and run_uid in run_runtime_resources_map.get(
project, {}
):
# if found resource there is no need to continue
return
last_update_str = run.get("status", {}).get("last_update")
debounce_period = (
config.resolve_runs_monitoring_missing_runtime_resources_debouncing_interval()
)
if last_update_str is None:
logger.info(
"Runs monitoring found run in non-terminal state without last update time set, "
"updating last update time to now, to be able to evaluate next time if something changed",
project=project,
uid=run_uid,
db_run_state=db_run_state,
now=now,
debounce_period=debounce_period,
)
run.setdefault("status", {})["last_update"] = now.isoformat()
db.store_run(db_session, run, run_uid, project)
return
if datetime.fromisoformat(last_update_str) > now - timedelta(
seconds=debounce_period
):
# we are setting non-terminal states to runs before the run is actually applied to k8s, meaning there is
# a timeframe where the run exists and no runtime resources exist and it's ok, therefore we're applying
# a debounce period before setting the state to error
logger.warning(
"Monitoring did not discover a runtime resource that corresponded to a run in a "
"non-terminal state. but record has recently updated. Debouncing",
project=project,
uid=run_uid,
db_run_state=db_run_state,
last_update=datetime.fromisoformat(last_update_str),
now=now,
debounce_period=debounce_period,
)
else:
logger.info(
"Updating run state", run_uid=run_uid, run_state=RunStates.error
)
run.setdefault("status", {})["state"] = RunStates.error
run.setdefault("status", {})["last_update"] = now.isoformat()
db.store_run(db_session, run, run_uid, project)
def _add_object_label_selector_if_needed(
self,
object_id: typing.Optional[str] = None,
label_selector: typing.Optional[str] = None,
):
if object_id:
object_label_selector = self._get_object_label_selector(object_id)
if label_selector:
label_selector = ",".join([object_label_selector, label_selector])
else:
label_selector = object_label_selector
return label_selector
def _enrich_list_resources_response(
self,
response: Union[
mlrun.api.schemas.RuntimeResources,
mlrun.api.schemas.GroupedByJobRuntimeResourcesOutput,
mlrun.api.schemas.GroupedByProjectRuntimeResourcesOutput,
],
namespace: str,
label_selector: str = None,
group_by: Optional[mlrun.api.schemas.ListRuntimeResourcesGroupByField] = None,
) -> Union[
mlrun.api.schemas.RuntimeResources,
mlrun.api.schemas.GroupedByJobRuntimeResourcesOutput,
mlrun.api.schemas.GroupedByProjectRuntimeResourcesOutput,
]:
"""
Override this to list resources other then pods or CRDs (which are handled by the base class)
"""
return response
def _build_output_from_runtime_resources(
self,
response: Union[
mlrun.api.schemas.RuntimeResources,
mlrun.api.schemas.GroupedByJobRuntimeResourcesOutput,
mlrun.api.schemas.GroupedByProjectRuntimeResourcesOutput,
],
runtime_resources_list: List[mlrun.api.schemas.RuntimeResources],
group_by: Optional[mlrun.api.schemas.ListRuntimeResourcesGroupByField] = None,
):
"""
Override this to add runtime resources other then pods or CRDs (which are handled by the base class) to the
output
"""
return response
def _delete_resources(
self,
db: DBInterface,
db_session: Session,
namespace: str,
deleted_resources: List[Dict],
label_selector: str = None,
force: bool = False,
grace_period: int = None,
):
"""
Override this to handle deletion of resources other then pods or CRDs (which are handled by the base class)
Note that this is happening before the deletion of the CRDs or the pods
Note to add this at the beginning:
if grace_period is None:
grace_period = config.runtime_resources_deletion_grace_period
"""
pass
def _resolve_crd_object_status_info(
self, db: DBInterface, db_session: Session, crd_object
) -> Tuple[bool, Optional[datetime], Optional[str]]:
"""
Override this if the runtime has CRD resources.
:return: Tuple with:
1. bool determining whether the crd object is in terminal state
2. datetime of when the crd object got into terminal state (only when the crd object in terminal state)
3. the desired run state matching the crd object state
"""
return False, None, None
def _update_ui_url(
self,
db: DBInterface,
db_session: Session,
project: str,
uid: str,
crd_object,
run: Dict = None,
):
"""
Update the UI URL for relevant jobs.
"""
pass
def _resolve_pod_status_info(
self, db: DBInterface, db_session: Session, pod: Dict
) -> Tuple[bool, Optional[datetime], Optional[str]]:
"""
:return: Tuple with:
1. bool determining whether the pod is in terminal state
2. datetime of when the pod got into terminal state (only when the pod in terminal state)
3. the run state matching the pod state
"""
in_terminal_state = pod["status"]["phase"] in PodPhases.terminal_phases()
run_state = PodPhases.pod_phase_to_run_state(pod["status"]["phase"])
last_container_completion_time = None
if in_terminal_state:
for container_status in pod["status"].get("container_statuses", []):
if container_status.get("state", {}).get("terminated"):
container_completion_time = container_status["state"][
"terminated"
].get("finished_at")
# take latest completion time
if (
not last_container_completion_time
or last_container_completion_time < container_completion_time
):
last_container_completion_time = container_completion_time
return in_terminal_state, last_container_completion_time, run_state
def _get_default_label_selector(self) -> str:
"""
Override this to add a default label selector
"""
class_values = self._get_possible_mlrun_class_label_values()
if not class_values:
return ""
if len(class_values) == 1:
return f"mlrun/class={class_values[0]}"
return f"mlrun/class in ({', '.join(class_values)})"
@staticmethod
def _get_crd_info() -> Tuple[str, str, str]:
"""
Override this if the runtime has CRD resources. this should return the CRD info:
crd group, crd version, crd plural
"""
return "", "", ""
@staticmethod
def _are_resources_coupled_to_run_object() -> bool:
"""
Some resources are tightly coupled to mlrun Run object, for example, for each Run of a Function of the job kind
a kubernetes job is being generated, on the opposite a Function of the daskjob kind generates a dask cluster,
and every Run is being executed using this cluster, i.e. no resources are created for the Run.
This function should return true for runtimes in which Run are coupled to the underlying resources and therefore
aspects of the Run (like its state) should be taken into consideration on resources deletion
"""
return False
@staticmethod
def _expect_pods_without_uid() -> bool:
return False
def _list_pods(self, namespace: str, label_selector: str = None) -> List:
k8s_helper = get_k8s_helper()
pods = k8s_helper.list_pods(namespace, selector=label_selector)
# when we work with custom objects (list_namespaced_custom_object) it's always a dict, to be able to generalize
# code working on runtime resource (either a custom object or a pod) we're transforming to dicts
pods = [pod.to_dict() for pod in pods]
return pods
def _list_crd_objects(self, namespace: str, label_selector: str = None) -> List:
k8s_helper = get_k8s_helper()
crd_group, crd_version, crd_plural = self._get_crd_info()
crd_objects = []
if crd_group and crd_version and crd_plural:
try:
crd_objects = k8s_helper.crdapi.list_namespaced_custom_object(
crd_group,
crd_version,
namespace,
crd_plural,
label_selector=label_selector,
)
except ApiException as exc:
# ignore error if crd is not defined
if exc.status != 404:
raise
else:
crd_objects = crd_objects["items"]
return crd_objects
def _resolve_label_selector(
self,
project: str,
object_id: typing.Optional[str] = None,
label_selector: typing.Optional[str] = None,
) -> str:
default_label_selector = self._get_default_label_selector()
if label_selector:
label_selector = ",".join([default_label_selector, label_selector])
else:
label_selector = default_label_selector
if project and project != "*":
label_selector = ",".join([label_selector, f"mlrun/project={project}"])
label_selector = self._add_object_label_selector_if_needed(
object_id, label_selector
)
return label_selector
def _wait_for_pods_deletion(
self,
namespace: str,
deleted_pods: List[Dict],
label_selector: str = None,
):
k8s_helper = get_k8s_helper()
deleted_pod_names = [pod_dict["metadata"]["name"] for pod_dict in deleted_pods]
def _verify_pods_removed():
pods = k8s_helper.v1api.list_namespaced_pod(
namespace, label_selector=label_selector
)
existing_pod_names = [pod.metadata.name for pod in pods.items]
still_in_deletion_pods = set(existing_pod_names).intersection(
deleted_pod_names
)
if still_in_deletion_pods:
raise RuntimeError(
f"Pods are still in deletion process: {still_in_deletion_pods}"
)
if deleted_pod_names:
timeout = 180
logger.debug(
"Waiting for pods deletion",
timeout=timeout,
interval=self.wait_for_deletion_interval,
)
mlrun.utils.retry_until_successful(
self.wait_for_deletion_interval,
timeout,
logger,
True,
_verify_pods_removed,
)
def _wait_for_crds_underlying_pods_deletion(
self,
deleted_crds: List[Dict],
label_selector: str = None,
):
# we're using here the run identifier as the common ground to identify which pods are relevant to which CRD, so
# if they are not coupled we are not able to wait - simply return
# NOTE - there are surely smarter ways to do this, without depending on the run object, but as of writing this
# none of the runtimes using CRDs are like that, so not handling it now
if not self._are_resources_coupled_to_run_object():
return
def _verify_crds_underlying_pods_removed():
project_uid_crd_map = {}
for crd in deleted_crds:
project, uid, _ = self._resolve_runtime_resource_run(crd)
if not uid or not project:
logger.warning(
"Could not resolve run uid from crd. Skipping waiting for pods deletion",
crd=crd,
)
continue
project_uid_crd_map.setdefault(project, {})[uid] = crd["metadata"][
"name"
]
still_in_deletion_crds_to_pod_names = {}
jobs_runtime_resources: mlrun.api.schemas.GroupedByJobRuntimeResourcesOutput = self.list_resources(
"*",
label_selector=label_selector,
group_by=mlrun.api.schemas.ListRuntimeResourcesGroupByField.job,
)
for project, project_jobs in jobs_runtime_resources.items():
if project not in project_uid_crd_map:
continue
for job_uid, job_runtime_resources in jobs_runtime_resources[
project
].items():
if job_uid not in project_uid_crd_map[project]:
continue
if job_runtime_resources.pod_resources:
still_in_deletion_crds_to_pod_names[
project_uid_crd_map[project][job_uid]
] = [
pod_resource.name
for pod_resource in job_runtime_resources.pod_resources
]
if still_in_deletion_crds_to_pod_names:
raise RuntimeError(
f"CRD underlying pods are still in deletion process: {still_in_deletion_crds_to_pod_names}"
)
if deleted_crds:
timeout = 180
logger.debug(
"Waiting for CRDs underlying pods deletion",
timeout=timeout,
interval=self.wait_for_deletion_interval,
)
mlrun.utils.retry_until_successful(
self.wait_for_deletion_interval,
timeout,
logger,
True,
_verify_crds_underlying_pods_removed,
)
def _delete_pod_resources(
self,
db: DBInterface,
db_session: Session,
namespace: str,
label_selector: str = None,
force: bool = False,
grace_period: int = None,
) -> List[Dict]:
if grace_period is None:
grace_period = config.runtime_resources_deletion_grace_period
k8s_helper = get_k8s_helper()
pods = k8s_helper.v1api.list_namespaced_pod(
namespace, label_selector=label_selector
)
deleted_pods = []
for pod in pods.items:
pod_dict = pod.to_dict()
# best effort - don't let one failure in pod deletion to cut the whole operation
try:
(
in_terminal_state,
last_update,
run_state,
) = self._resolve_pod_status_info(db, db_session, pod_dict)
if not force:
if not in_terminal_state:
continue
# give some grace period if we have last update time
now = datetime.now(timezone.utc)
if (
last_update is not None
and last_update + timedelta(seconds=float(grace_period)) > now
):
continue
# if resources are tightly coupled to the run object - we want to perform some actions on the run object
# before deleting them
if self._are_resources_coupled_to_run_object():
try:
self._pre_deletion_runtime_resource_run_actions(
db, db_session, pod_dict, run_state
)
except Exception as exc:
# Don't prevent the deletion for failure in the pre deletion run actions
logger.warning(
"Failure in pod run pre-deletion actions. Continuing",
exc=repr(exc),
pod_name=pod.metadata.name,
)
self._delete_pod(namespace, pod)
deleted_pods.append(pod_dict)
except Exception as exc:
logger.warning(
f"Cleanup failed processing pod {pod.metadata.name}: {repr(exc)}. Continuing"
)
self._wait_for_pods_deletion(namespace, deleted_pods, label_selector)
return deleted_pods
def _delete_crd_resources(
self,
db: DBInterface,
db_session: Session,
namespace: str,
label_selector: str = None,
force: bool = False,
grace_period: int = None,
) -> List[Dict]:
if grace_period is None:
grace_period = config.runtime_resources_deletion_grace_period
k8s_helper = get_k8s_helper()
crd_group, crd_version, crd_plural = self._get_crd_info()
deleted_crds = []
try:
crd_objects = k8s_helper.crdapi.list_namespaced_custom_object(
crd_group,
crd_version,
namespace,
crd_plural,
label_selector=label_selector,
)
except ApiException as exc:
# ignore error if crd is not defined
if exc.status != 404:
raise
else:
for crd_object in crd_objects["items"]:
# best effort - don't let one failure in pod deletion to cut the whole operation
try:
(
in_terminal_state,
last_update,
desired_run_state,
) = self._resolve_crd_object_status_info(db, db_session, crd_object)
if not force:
if not in_terminal_state:
continue
# give some grace period if we have last update time
now = datetime.now(timezone.utc)
if (
last_update is not None
and last_update + timedelta(seconds=float(grace_period))
> now
):
continue
# if resources are tightly coupled to the run object - we want to perform some actions on the run
# object before deleting them
if self._are_resources_coupled_to_run_object():
try:
self._pre_deletion_runtime_resource_run_actions(
db,
db_session,
crd_object,
desired_run_state,
)
except Exception as exc:
# Don't prevent the deletion for failure in the pre deletion run actions
logger.warning(
"Failure in crd object run pre-deletion actions. Continuing",
exc=str(exc),
crd_object_name=crd_object["metadata"]["name"],
)
self._delete_crd(
namespace, crd_group, crd_version, crd_plural, crd_object
)
deleted_crds.append(crd_object)
except Exception:
exc = traceback.format_exc()
crd_object_name = crd_object["metadata"]["name"]
logger.warning(
f"Cleanup failed processing CRD object {crd_object_name}: {exc}. Continuing"
)
self._wait_for_crds_underlying_pods_deletion(deleted_crds, label_selector)
return deleted_crds
def _pre_deletion_runtime_resource_run_actions(
self,
db: DBInterface,
db_session: Session,
runtime_resource: Dict,
run_state: str,
):
project, uid, name = self._resolve_runtime_resource_run(runtime_resource)
# if cannot resolve related run nothing to do
if not uid:
if not self._expect_pods_without_uid():
logger.warning(
"Could not resolve run uid from runtime resource. Skipping pre-deletion actions",
runtime_resource=runtime_resource,
)
raise ValueError("Could not resolve run uid from runtime resource")
else:
return
logger.info(
"Performing pre-deletion actions before cleaning up runtime resources",
project=project,
uid=uid,
)
self._ensure_run_state(db, db_session, project, uid, name, run_state)
self._ensure_run_logs_collected(db, db_session, project, uid)
def _is_runtime_resource_run_in_terminal_state(
self,
db: DBInterface,
db_session: Session,
runtime_resource: Dict,
) -> Tuple[bool, Optional[datetime]]:
"""
A runtime can have different underlying resources (like pods or CRDs) - to generalize we call it runtime
resource. This function will verify whether the Run object related to this runtime resource is in transient
state. This is useful in order to determine whether an object can be removed. for example, a kubejob's pod
might be in completed state, but we would like to verify that the run is completed as well to verify the logs
were collected before we're removing the pod.
:returns: bool determining whether the run in terminal state, and the last update time if it exists
"""
project, uid, _ = self._resolve_runtime_resource_run(runtime_resource)
# if no uid, assume in terminal state
if not uid:
return True, None
run = db.read_run(db_session, uid, project)
last_update = None
last_update_str = run.get("status", {}).get("last_update")
if last_update_str is not None:
last_update = datetime.fromisoformat(last_update_str)
if run.get("status", {}).get("state") not in RunStates.terminal_states():
return False, last_update
return True, last_update
def _list_runs_for_monitoring(
self, db: DBInterface, db_session: Session, states: list = None
):
runs = db.list_runs(db_session, project="*", states=states)
project_run_uid_map = {}
run_with_missing_data = []
duplicated_runs = []
for run in runs:
project = run.get("metadata", {}).get("project")
uid = run.get("metadata", {}).get("uid")
if not uid or not project:
run_with_missing_data.append(run.get("metadata", {}))
continue
current_run = project_run_uid_map.setdefault(project, {}).get(uid)
# sanity
if current_run:
duplicated_runs = {
"monitored_run": current_run.get(["metadata"]),
"duplicated_run": run.get(["metadata"]),
}
continue
project_run_uid_map[project][uid] = run
# If there are duplications or runs with missing data it probably won't be fixed
# Monitoring is running periodically and we don't want to log on every problem we found which will spam the log
# so we're aggregating the problems and logging only once per aggregation
if duplicated_runs:
logger.warning(
"Found duplicated runs (same uid). Heuristically monitoring the first one found",
duplicated_runs=duplicated_runs,
)
if run_with_missing_data:
logger.warning(
"Found runs with missing data. They will not be monitored",
run_with_missing_data=run_with_missing_data,
)
return project_run_uid_map
def _monitor_runtime_resource(
self,
db: DBInterface,
db_session: Session,
project_run_uid_map: Dict,
runtime_resource: Dict,
runtime_resource_is_crd: bool,
namespace: str,
project: str = None,
uid: str = None,
name: str = None,
):
if not project and not uid and not name:
project, uid, name = self._resolve_runtime_resource_run(runtime_resource)
if not project or not uid:
# Currently any build pod won't have UID and therefore will cause this log message to be printed which
# spams the log
# TODO: uncomment the log message when builder become a kind / starts having a UID
# logger.warning(
# "Could not resolve run project or uid from runtime resource, can not monitor run. Continuing",
# project=project,
# uid=uid,
# runtime_resource_name=runtime_resource["metadata"]["name"],
# namespace=namespace,
# )
return
run = project_run_uid_map.get(project, {}).get(uid)
if runtime_resource_is_crd:
(
_,
_,
run_state,
) = self._resolve_crd_object_status_info(db, db_session, runtime_resource)
else:
(
_,
_,
run_state,
) = self._resolve_pod_status_info(db, db_session, runtime_resource)
self._update_ui_url(db, db_session, project, uid, runtime_resource, run)
_, updated_run_state = self._ensure_run_state(
db,
db_session,
project,
uid,
name,
run_state,
run,
search_run=False,
)
if updated_run_state in RunStates.terminal_states():
self._ensure_run_logs_collected(db, db_session, project, uid)
def _build_list_resources_response(
self,
pod_resources: List[mlrun.api.schemas.RuntimeResource] = None,
crd_resources: List[mlrun.api.schemas.RuntimeResource] = None,
group_by: Optional[mlrun.api.schemas.ListRuntimeResourcesGroupByField] = None,
) -> Union[
mlrun.api.schemas.RuntimeResources,
mlrun.api.schemas.GroupedByJobRuntimeResourcesOutput,
mlrun.api.schemas.GroupedByProjectRuntimeResourcesOutput,
]:
if crd_resources is None:
crd_resources = []
if pod_resources is None:
pod_resources = []
if group_by is None:
return mlrun.api.schemas.RuntimeResources(
crd_resources=crd_resources, pod_resources=pod_resources
)
else:
if group_by == mlrun.api.schemas.ListRuntimeResourcesGroupByField.job:
return self._build_grouped_by_job_list_resources_response(
pod_resources, crd_resources
)
elif group_by == mlrun.api.schemas.ListRuntimeResourcesGroupByField.project:
return self._build_grouped_by_project_list_resources_response(
pod_resources, crd_resources
)
else:
raise NotImplementedError(
f"Provided group by field is not supported. group_by={group_by}"
)
def _build_grouped_by_project_list_resources_response(
self,
pod_resources: List[mlrun.api.schemas.RuntimeResource] = None,
crd_resources: List[mlrun.api.schemas.RuntimeResource] = None,
) -> mlrun.api.schemas.GroupedByProjectRuntimeResourcesOutput:
resources = {}
for pod_resource in pod_resources:
self._add_resource_to_grouped_by_project_resources_response(
resources, "pod_resources", pod_resource
)
for crd_resource in crd_resources:
self._add_resource_to_grouped_by_project_resources_response(
resources, "crd_resources", crd_resource
)
return resources
def _build_grouped_by_job_list_resources_response(
self,
pod_resources: List[mlrun.api.schemas.RuntimeResource] = None,
crd_resources: List[mlrun.api.schemas.RuntimeResource] = None,
) -> mlrun.api.schemas.GroupedByJobRuntimeResourcesOutput:
resources = {}
for pod_resource in pod_resources:
self._add_resource_to_grouped_by_job_resources_response(
resources, "pod_resources", pod_resource
)
for crd_resource in crd_resources:
self._add_resource_to_grouped_by_job_resources_response(
resources, "crd_resources", crd_resource
)
return resources
def _add_resource_to_grouped_by_project_resources_response(
self,
resources: mlrun.api.schemas.GroupedByJobRuntimeResourcesOutput,
resource_field_name: str,
resource: mlrun.api.schemas.RuntimeResource,
):
if "mlrun/class" in resource.labels:
project = resource.labels.get("mlrun/project", "")
mlrun_class = resource.labels["mlrun/class"]
kind = self._resolve_kind_from_class(mlrun_class)
self._add_resource_to_grouped_by_field_resources_response(
project, kind, resources, resource_field_name, resource
)
def _add_resource_to_grouped_by_job_resources_response(
self,
resources: mlrun.api.schemas.GroupedByJobRuntimeResourcesOutput,
resource_field_name: str,
resource: mlrun.api.schemas.RuntimeResource,
):
if "mlrun/uid" in resource.labels:
project = resource.labels.get("mlrun/project", config.default_project)
uid = resource.labels["mlrun/uid"]
self._add_resource_to_grouped_by_field_resources_response(
project, uid, resources, resource_field_name, resource
)
@staticmethod
def _add_resource_to_grouped_by_field_resources_response(
first_field_value: str,
second_field_value: str,
resources: mlrun.api.schemas.GroupedByJobRuntimeResourcesOutput,
resource_field_name: str,
resource: mlrun.api.schemas.RuntimeResource,
):
if first_field_value not in resources:
resources[first_field_value] = {}
if second_field_value not in resources[first_field_value]:
resources[first_field_value][
second_field_value
] = mlrun.api.schemas.RuntimeResources(pod_resources=[], crd_resources=[])
if not getattr(
resources[first_field_value][second_field_value], resource_field_name
):
setattr(
resources[first_field_value][second_field_value],
resource_field_name,
[],
)
getattr(
resources[first_field_value][second_field_value], resource_field_name
).append(resource)
@staticmethod
def _resolve_kind_from_class(mlrun_class: str) -> str:
class_to_kind_map = {}
for kind in mlrun.runtimes.RuntimeKinds.runtime_with_handlers():
runtime_handler = mlrun.runtimes.get_runtime_handler(kind)
class_values = runtime_handler._get_possible_mlrun_class_label_values()
for value in class_values:
class_to_kind_map[value] = kind
return class_to_kind_map[mlrun_class]
@staticmethod
def _get_run_label_selector(project: str, run_uid: str):
return f"mlrun/project={project},mlrun/uid={run_uid}"
@staticmethod
def _ensure_run_logs_collected(
db: DBInterface, db_session: Session, project: str, uid: str
):
# import here to avoid circular imports
import mlrun.api.crud as crud
log_file_exists = crud.Logs().log_file_exists(project, uid)
if not log_file_exists:
_, logs_from_k8s = crud.Logs().get_logs(
db_session, project, uid, source=LogSources.K8S
)
if logs_from_k8s:
logger.info("Storing run logs", project=project, uid=uid)
crud.Logs().store_log(logs_from_k8s, project, uid, append=False)
@staticmethod
def _ensure_run_state(
db: DBInterface,
db_session: Session,
project: str,
uid: str,
name: str,
run_state: str,
run: Dict = None,
search_run: bool = True,
) -> Tuple[bool, str]:
if run is None:
run = {}
if search_run:
try:
run = db.read_run(db_session, uid, project)
except mlrun.errors.MLRunNotFoundError:
run = {}
if not run:
logger.warning(
"Run not found. A new run will be created",
project=project,
uid=uid,
desired_run_state=run_state,
search_run=search_run,
)
run = {"metadata": {"project": project, "name": name, "uid": uid}}
db_run_state = run.get("status", {}).get("state")
if db_run_state:
if db_run_state == run_state:
return False, run_state
# if the current run state is terminal and different than the desired - log
if db_run_state in RunStates.terminal_states():
# This can happen when the SDK running in the user's Run updates the Run's state to terminal, but
# before it exits, when the runtime resource is still running, the API monitoring (here) is executed
if run_state not in RunStates.terminal_states():
now = datetime.now(timezone.utc)
last_update_str = run.get("status", {}).get("last_update")
if last_update_str is not None:
last_update = datetime.fromisoformat(last_update_str)
debounce_period = config.runs_monitoring_interval
if last_update > now - timedelta(
seconds=float(debounce_period)
):
logger.warning(
"Monitoring found non-terminal state on runtime resource but record has recently "
"updated to terminal state. Debouncing",
project=project,
uid=uid,
db_run_state=db_run_state,
run_state=run_state,
last_update=last_update,
now=now,
debounce_period=debounce_period,
)
return False, run_state
logger.warning(
"Run record has terminal state but monitoring found different state on runtime resource. Changing",
project=project,
uid=uid,
db_run_state=db_run_state,
run_state=run_state,
)
logger.info("Updating run state", run_state=run_state)
run.setdefault("status", {})["state"] = run_state
run.setdefault("status", {})["last_update"] = now_date().isoformat()
db.store_run(db_session, run, uid, project)
return True, run_state
@staticmethod
def _resolve_runtime_resource_run(runtime_resource: Dict) -> Tuple[str, str, str]:
project = (
runtime_resource.get("metadata", {}).get("labels", {}).get("mlrun/project")
)
if not project:
project = config.default_project
uid = runtime_resource.get("metadata", {}).get("labels", {}).get("mlrun/uid")
name = (
runtime_resource.get("metadata", {})
.get("labels", {})
.get("mlrun/name", "no-name")
)
return project, uid, name
@staticmethod
def _delete_crd(namespace, crd_group, crd_version, crd_plural, crd_object):
k8s_helper = get_k8s_helper()
name = crd_object["metadata"]["name"]
try:
k8s_helper.crdapi.delete_namespaced_custom_object(
crd_group,
crd_version,
namespace,
crd_plural,
name,
)
logger.info(
"Deleted crd object",
name=name,
namespace=namespace,
crd_plural=crd_plural,
)
except ApiException as exc:
# ignore error if crd object is already removed
if exc.status != 404:
raise
@staticmethod
def _delete_pod(namespace, pod):
k8s_helper = get_k8s_helper()
try:
k8s_helper.v1api.delete_namespaced_pod(pod.metadata.name, namespace)
logger.info("Deleted pod", pod=pod.metadata.name)
except ApiException as exc:
# ignore error if pod is already removed
if exc.status != 404:
raise
@staticmethod
def _build_pod_resources(pods) -> List[mlrun.api.schemas.RuntimeResource]:
pod_resources = []
for pod in pods:
pod_resources.append(
mlrun.api.schemas.RuntimeResource(
name=pod["metadata"]["name"],
labels=pod["metadata"]["labels"],
status=pod["status"],
)
)
return pod_resources
@staticmethod
def _build_crd_resources(custom_objects) -> List[mlrun.api.schemas.RuntimeResource]:
crd_resources = []
for custom_object in custom_objects:
crd_resources.append(
mlrun.api.schemas.RuntimeResource(
name=custom_object["metadata"]["name"],
labels=custom_object["metadata"]["labels"],
status=custom_object.get("status", {}),
)
)
return crd_resources
