Skip to content

Commit

Permalink
Add and expose settings
Browse files Browse the repository at this point in the history
  • Loading branch information
@chicocvenancio
chicocvenancio committed Dec 13, 2024
1 parent bcef3f9 commit 4e32467
Show file tree
Hide file tree
Showing 2 changed files with 48 additions and 14 deletions.
45 changes: 37 additions & 8 deletions argyle_recommender/recommender.py
View file
Original file line number Diff line number Diff line change
Expand Up @@ -82,7 +82,8 @@ def _format(value):
return resource_units.format(value)


def get_krr_json(label, namespace="*", prometheus=None, context=None, app_name=None) -> Result:
def get_krr_json(label, namespace="*", prometheus=None, context=None, app_name=None,
settings: dict = dict()) -> Result:
log.debug("loading recomendations for %s in ns %s", label, namespace)
if label == "no-selector":
selector = []
Expand All @@ -105,6 +106,15 @@ def get_krr_json(label, namespace="*", prometheus=None, context=None, app_name=N
command.extend(namespace_flag)
command.extend(prometheus_flag)
command.extend(context_flag)
cpu_min = settings.get("cpu_min", 10)
mem_min = settings.get("mem_min", 100)
command.extend(["--cpu-min", str(cpu_min), "--mem-min", str(mem_min)])
cpu_bump_percentage = settings.get("cpu_bump_percentage", 20)
command.extend(["--cpu-bump-percentage", str(cpu_bump_percentage)])
secondary_cpu_percentile = settings.get("secondary_cpu_percentile", 50)
command.extend(["--secondary-cpu-percentile", str(secondary_cpu_percentile)])
memory_buffer_percentage = settings.get("memory_buffer_percentage", 120)
command.extend(["--memory-buffer-percentage", str(memory_buffer_percentage)])
log.info("%s: %s", app_name, ' '.join(command))
krr = subprocess.run(command, capture_output=True, encoding="utf8", check=True).stdout
results = None
Expand Down Expand Up @@ -141,13 +151,15 @@ def get_krr_json(label, namespace="*", prometheus=None, context=None, app_name=N
return results


def find_recommendations(build_yaml_path, namespace_flag, prometheus, context=None):
def find_recommendations(build_yaml_path, namespace_flag, prometheus, context=None,
settings: dict = dict()) -> Optional[Result]:
scans = []
results = None
if build_yaml_path == "no-selector":
namespace = namespace_flag
return get_krr_json("no-selector", namespace=namespace,
prometheus=prometheus, context=context)
prometheus=prometheus, context=context,
settings=settings)

with open(build_yaml_path, "r", encoding="utf-8") as build_file:
yaml = YAML()
Expand Down Expand Up @@ -183,7 +195,7 @@ def find_recommendations(build_yaml_path, namespace_flag, prometheus, context=No
namespace = namespace_flag
results = get_krr_json(
f"{label_name}={label}", namespace, prometheus,
app_name=app_name, context=context)
app_name=app_name, context=context, settings=settings)
if len(results.scans) == 0:
docs.pop(0)
continue
Expand Down Expand Up @@ -490,13 +502,15 @@ def handle_kustomizations(resources_path: str):
def process_app(path: Union[str, pathlib.Path], namespace,
create_pr_flag: Union[bool, None] = False,
prometheus=None, __key=None, context=None,
clean_resources_flag: bool = False):
clean_resources_flag: bool = False,
settings: dict = dict()):
if create_pr_flag is None:
create_pr_flag = False

repo = pull_repo(__key)
os.chdir(repo.working_dir)
results = find_recommendations(path, namespace, prometheus, context=context)
results = find_recommendations(path, namespace, prometheus, context=context,
settings=settings)
if results is None:
return None
if clean_resources_flag:
Expand Down Expand Up @@ -665,8 +679,14 @@ def main(path: Annotated[str, typer.Argument],
incluster: Annotated[Optional[bool], typer.Option("--incluster")] = False,
clean_resources_flag: Annotated[bool, typer.Option("--clean-resources")] = False,
debug: Annotated[bool, typer.Option("--debug")] = False,
cpu_min: Annotated[int, typer.Option("--cpu-min")] = 10,
mem_min: Annotated[int, typer.Option("--mem-min")] = 100,
cpu_bump_percentage: Annotated[float, typer.Option("--cpu-bump-percentage")] = 20,
secondary_cpu_percentile: Annotated[float, typer.Option("--secondary-cpu-percentile")] = 50,
memory_buffer_percentage: Annotated[float, typer.Option("--memory-buffer-percentage")] = 120,
):


key_file = os.environ.get("PRIVATE_KEY_FILE")
if key_file:
with open(key_file, "r", encoding="utf-8") as _f:
Expand All @@ -681,6 +701,14 @@ def main(path: Annotated[str, typer.Argument],
" path or PRIVATE_KEY with the key contents.")
repo = pull_repo(__key)
os.chdir(repo.working_dir)

settings = dict(
cpu_bump_percentage=cpu_bump_percentage,
memory_buffer_percentage=memory_buffer_percentage,
secondary_cpu_percentile=secondary_cpu_percentile,
cpu_min=cpu_min,
mem_min=mem_min,
)
if path in ["prod", "all"] or path.startswith(".build"):
if path.startswith(".build"):
build_yamls(path)
Expand All @@ -691,12 +719,13 @@ def main(path: Annotated[str, typer.Argument],
context = app_yaml.name.replace(".yaml", "").split("-")[-1]
try:
process_app(app_yaml, namespace, create_pr, prometheus, __key, context=context,
clean_resources_flag=clean_resources_flag)
clean_resources_flag=clean_resources_flag, settings=settings
)
except Exception:
log.exception("error in processing app %s", app_yaml, stack_info=True)
else:
process_app(path, namespace, create_pr, prometheus, __key, context=context,
clean_resources_flag=clean_resources_flag)
clean_resources_flag=clean_resources_flag, settings=settings)


if __name__ == "__main__":
Expand Down
17 changes: 11 additions & 6 deletions robusta_krr/strategies/krr_argyle_strategy.py
View file
Original file line number Diff line number Diff line change
Expand Up @@ -57,6 +57,7 @@ def get_query(self, object: K8sObjectData, duration: str, step: str) -> str:
return SecondaryPercentileCPULoader


#TODO move to MaxOOMKilledMemoryLoader
class OOMKilledLoader(PrometheusMetric):
"""
A metric loader for loading OOMKilled metrics.
Expand All @@ -81,20 +82,24 @@ def get_query(self, object: K8sObjectData, duration: str, step: str) -> str:
# Providing description to the settings will make it available in the CLI help
class ArgyleStrategySettings(SimpleStrategySettings):
memory_buffer_percentage: float = pd.Field(
50, gt=0, description="The percentage of added buffer to the peak memory usage for memory limit recommendation."
120, gt=0, description="The percentage of added buffer to the peak memory usage for memory limit recommendation."
)

def calculate_memory_usage(self, data: PodsTimeData) -> float:
data_ = [np.max(values[:, 1]) for values in data.values()]
if len(data_) == 0:
return float("NaN")

return np.max(data_)
return float(np.max(data_))

secondary_cpu_percentile: float = pd.Field(
50, gt=0, description="The percentile to use for the secondary CPU usage metric."
)

cpu_bump_percentage: float = pd.Field(
20, gt=0, description="The percentage of added buffer to the CPU request."
)


class ArgyleStrategy(BaseStrategy[ArgyleStrategySettings]):
"""
Expand All @@ -112,7 +117,7 @@ def description(self):
History: {self.settings.history_duration} hours
Step: {self.settings.timeframe_duration} minutes
All parameters can be customized. For example: `krr simple --cpu_percentile=90 --memory_buffer_percentage=15 --history_duration=24 --timeframe_duration=0.5`
All parameters can be customized. For example: `krr argyle --cpu_percentile=90 --memory_buffer_percentage=15 --history_duration=24 --timeframe_duration=0.5`
""")
@staticmethod
def info_from_list(info_list: list[str]):
Expand Down Expand Up @@ -172,7 +177,7 @@ def __calculate_cpu_proposal(
# if object_data.labels:
# print(object_data.labels)

cpu_usage = self.settings.calculate_cpu_proposal(filtered_data)
cpu_usage = self.settings.calculate_cpu_proposal(filtered_data) * (1 + self.settings.cpu_bump_percentage / 100)
cpu_limit = (self._calculate_cpu_limit(cpu_usage))
return ResourceRecommendation(request=cpu_usage, limit=cpu_limit, info=info)

Expand All @@ -184,7 +189,7 @@ def __calculate_memory_proposal(
if len(data) == 0:
return ResourceRecommendation.undefined(info="No data")

data_count = {pod: values[0, 1] for pod, values in history_data["MemoryAmountLoader"].items()}
data_count = {pod: values[0, 1] for pod, values in data.items()}
# Here we filter out pods from calculation that have less than `points_required` data points
filtered_data = {
pod: value for pod, value in data.items() if data_count.get(pod, 0) >= self.settings.points_required
Expand All @@ -197,7 +202,7 @@ def __calculate_memory_proposal(
if object_data.hpa is not None and object_data.hpa.target_memory_utilization_percentage is not None:
info.append("Memory utilization HPA detected, be wary of ever increasing recommendations")


#TODO move to MaxOOMKilledMemoryLoader
if history_data.get("OOMKilledLoader", None):
info.append("Container has been OOMKilled in the period")
# if OOMKilled consider usage = limit
Expand Down

0 comments on commit 4e32467

Please sign in to comment.