unrip/scripts/ops/deployment_status.py
philipp aa8a868c14
All checks were successful
deploy / deploy (push) Successful in 41s
fix: OOM heuristic bug in quote outcomes
Proof: Successful trades only dashboard is empty
Assumptions: The downtime resulted in a massive gap we can attribute
Still fake: Some outcomes may be heuristic
2026-06-16 16:30:09 +02:00

250 lines
8.6 KiB
Python
Executable file

#!/usr/bin/env python3
from __future__ import annotations
import argparse
import json
import sys
from pathlib import Path
sys.path.insert(0, str(Path(__file__).resolve().parent))
from common import (
DEFAULT_NAMESPACE,
human_bytes,
human_duration,
kubectl_json,
now_utc,
parse_k8s_timestamp,
parse_storage_quantity,
print_table,
probe_path_usage,
)
def parse_args() -> argparse.Namespace:
parser = argparse.ArgumentParser(
description="Show current unrip deployment state, pod uptime, and mounted storage usage."
)
parser.add_argument(
"--namespace",
default=DEFAULT_NAMESPACE,
help=f"Kubernetes namespace to inspect (default: {DEFAULT_NAMESPACE})",
)
parser.add_argument(
"--include-rootfs",
action="store_true",
help="Also probe '/' for pods that do not have PVC-backed mounts.",
)
parser.add_argument(
"--include-completed",
action="store_true",
help="Include completed Job pods in the pod list.",
)
parser.add_argument(
"--output",
choices=("json", "table"),
default="json",
help="Output format (default: json).",
)
return parser.parse_args()
def ready_string(pod: dict) -> str:
statuses = pod.get("status", {}).get("containerStatuses", [])
ready = sum(1 for status in statuses if status.get("ready"))
total = len(statuses)
return f"{ready}/{total}"
def restart_count(pod: dict) -> int:
return sum(status.get("restartCount", 0) for status in pod.get("status", {}).get("containerStatuses", []))
def pvc_mounts_for_pod(pod: dict) -> list[dict[str, str]]:
volumes = {
volume["name"]: volume.get("persistentVolumeClaim", {}).get("claimName")
for volume in pod.get("spec", {}).get("volumes", [])
}
mounts: list[dict[str, str]] = []
for container in pod.get("spec", {}).get("containers", []):
for mount in container.get("volumeMounts", []):
claim_name = volumes.get(mount["name"])
if claim_name:
mounts.append(
{
"container": container.get("name", "app"),
"claim_name": claim_name,
"mount_path": mount["mountPath"],
}
)
return mounts
def is_completed_job_pod(pod: dict) -> bool:
owners = pod.get("metadata", {}).get("ownerReferences", [])
if not any(owner.get("kind") == "Job" for owner in owners):
return False
return pod.get("status", {}).get("phase") in {"Succeeded", "Failed"}
def main() -> int:
args = parse_args()
namespace = args.namespace
deployment_payload = kubectl_json("get", "deploy", "-n", namespace, "-o", "json")
pod_payload = kubectl_json("get", "pods", "-n", namespace, "-o", "json")
pvc_payload = kubectl_json("get", "pvc", "-n", namespace, "-o", "json")
pvc_requests = {
item["metadata"]["name"]: item.get("spec", {})
.get("resources", {})
.get("requests", {})
.get("storage", "")
for item in pvc_payload.get("items", [])
}
deployments: list[dict[str, object]] = []
for item in sorted(deployment_payload.get("items", []), key=lambda value: value["metadata"]["name"]):
container = item.get("spec", {}).get("template", {}).get("spec", {}).get("containers", [{}])[0]
deployments.append(
{
"name": item["metadata"]["name"],
"ready": {
"ready_replicas": item.get("status", {}).get("readyReplicas", 0),
"desired_replicas": item.get("spec", {}).get("replicas", 0),
"display": f"{item.get('status', {}).get('readyReplicas', 0)}/{item.get('spec', {}).get('replicas', 0)}",
},
"available_replicas": item.get("status", {}).get("availableReplicas", 0),
"image": container.get("image", "-"),
}
)
pods: list[dict[str, object]] = []
storage: list[dict[str, object]] = []
for pod in sorted(pod_payload.get("items", []), key=lambda value: value["metadata"]["name"]):
if not args.include_completed and is_completed_job_pod(pod):
continue
pod_name = pod["metadata"]["name"]
start_time = parse_k8s_timestamp(
pod.get("status", {}).get("startTime")
or pod.get("status", {}).get("containerStatuses", [{}])[0]
.get("state", {})
.get("running", {})
.get("startedAt")
)
uptime = human_duration((now_utc() - start_time).total_seconds() if start_time else None)
image = pod.get("spec", {}).get("containers", [{}])[0].get("image", "-")
pods.append(
{
"name": pod_name,
"phase": pod.get("status", {}).get("phase", "-"),
"ready": ready_string(pod),
"restarts": restart_count(pod),
"uptime": uptime,
"node": pod.get("spec", {}).get("nodeName", "-"),
"image": image,
}
)
mounts = pvc_mounts_for_pod(pod)
if not mounts and args.include_rootfs:
mounts = [{"claim_name": "-", "mount_path": "/", "container": "app"}]
if not mounts:
storage.append(
{
"pod": pod_name,
"path": None,
"claim": None,
"requested_bytes": None,
"requested": None,
"path_bytes": None,
"fs_used_bytes": None,
"fs_available_bytes": None,
"use_percent": None,
"note": "no pvc mounts",
}
)
continue
for mount in mounts:
usage = probe_path_usage(pod_name, mount["mount_path"], namespace=namespace)
requested_raw = pvc_requests.get(mount["claim_name"], "")
requested_bytes = parse_storage_quantity(requested_raw)
storage.append(
{
"pod": pod_name,
"path": mount["mount_path"],
"claim": mount["claim_name"],
"requested_bytes": requested_bytes,
"requested": requested_raw or None,
"path_bytes": usage["path_bytes"],
"fs_used_bytes": usage["filesystem_used_bytes"],
"fs_available_bytes": usage["filesystem_available_bytes"],
"use_percent": usage["filesystem_use_percent"],
"note": None,
}
)
if args.output == "json":
print(
json.dumps(
{
"namespace": namespace,
"deployments": deployments,
"pods": pods,
"storage": storage,
"notes": [
"storage figures come from container-mounted paths",
"pods without pvc-backed mounts do not report dedicated storage",
],
},
indent=2,
)
)
return 0
deployment_rows = [
[item["name"], item["ready"]["display"], str(item["available_replicas"]), item["image"]]
for item in deployments
]
pod_rows = [
[item["name"], item["phase"], item["ready"], str(item["restarts"]), item["uptime"], item["node"], item["image"]]
for item in pods
]
storage_rows = [
[
item["pod"],
item["path"] or "-",
item["claim"] or "no pvc mounts",
human_bytes(item["requested_bytes"]) if item["requested_bytes"] is not None else (item["requested"] or "-"),
human_bytes(item["path_bytes"]),
human_bytes(item["fs_used_bytes"]),
human_bytes(item["fs_available_bytes"]),
str(item["use_percent"] or "-"),
]
for item in storage
]
print(f"Namespace: {namespace}")
print()
print("Deployments")
print_table(["NAME", "READY", "AVAILABLE", "IMAGE"], deployment_rows)
print()
print("Pods")
print_table(["POD", "PHASE", "READY", "RESTARTS", "UPTIME", "NODE", "IMAGE"], pod_rows)
print()
print("Storage")
print_table(
["POD", "PATH", "CLAIM", "REQUESTED", "PATH_BYTES", "FS_USED", "FS_AVAIL", "USE%"],
storage_rows,
)
print()
print("Note: storage figures come from container-mounted paths. Pods without PVC-backed mounts report no dedicated storage.")
return 0
if __name__ == "__main__":
raise SystemExit(main())