release 0.0.0

This commit is contained in:
Waylon S. Walker 2025-03-24 14:53:16 -05:00
parent 24256b767d
commit 778f7fbd7c
8 changed files with 513 additions and 57 deletions

706
krayt.py Executable file
View file

@ -0,0 +1,706 @@
#!/usr/bin/env -S uv run --quiet --script
# /// script
# requires-python = ">=3.10"
# dependencies = [
# "typer",
# "kubernetes",
# "iterfzf"
# ]
# ///
"""
Krayt - The Kubernetes Volume Inspector
Like cracking open a Krayt dragon pearl, this tool helps you inspect what's inside your Kubernetes volumes.
Hunt down storage issues and explore your persistent data like a true Tatooine dragon hunter.
Features:
- Create inspector pods with all the tools you need
- Access volumes and device mounts from any pod
- Fuzzy search across all namespaces
- Built-in tools for file exploration and analysis
- Automatic cleanup of inspector pods
May the Force be with your volumes!
"""
from iterfzf import iterfzf
from kubernetes import client, config
import logging
import os
import time
import typer
from typing import Any, Optional
import yaml
KRAYT_VERSION = "NIGHTLY"
logging.basicConfig(level=logging.WARNING)
app = typer.Typer()
def clean_dict(d: dict[str, Any]) -> dict[str, Any]:
"""Remove None values and empty dicts from a dictionary recursively."""
if not isinstance(d, dict):
return d
return {
k: clean_dict(v)
for k, v in d.items()
if v is not None and v != {} and not (isinstance(v, dict) and not clean_dict(v))
}
def format_volume_mount(vm: client.V1VolumeMount) -> dict[str, Any]:
"""Format volume mount with only relevant fields."""
# Skip Kubernetes service account mounts
if vm.mount_path.startswith("/var/run/secrets/kubernetes.io/"):
return None
return clean_dict(
{
"name": vm.name,
"mountPath": vm.mount_path,
"readOnly": vm.read_only if vm.read_only else None,
}
)
def format_volume(v: client.V1Volume) -> dict[str, Any]:
"""Format volume into a dictionary, return None if it should be skipped"""
# Skip Kubernetes service account volumes
if v.name.startswith("kube-api-access-"):
return None
volume_source = None
if v.persistent_volume_claim:
volume_source = {
"persistentVolumeClaim": {"claimName": v.persistent_volume_claim.claim_name}
}
elif v.config_map:
volume_source = {"configMap": {"name": v.config_map.name}}
elif v.secret:
volume_source = {"secret": {"secretName": v.secret.secret_name}}
elif v.host_path: # Add support for hostPath volumes (used for device mounts)
volume_source = {
"hostPath": {
"path": v.host_path.path,
"type": v.host_path.type if v.host_path.type else None,
}
}
elif v.empty_dir: # Add support for emptyDir volumes (used for /dev/shm)
volume_source = {
"emptyDir": {
"medium": v.empty_dir.medium if v.empty_dir.medium else None,
"sizeLimit": v.empty_dir.size_limit if v.empty_dir.size_limit else None,
}
}
if not volume_source:
return None
return clean_dict({"name": v.name, **volume_source})
def fuzzy_select(items):
"""Use fzf to select from a list of (name, namespace) tuples"""
if not items:
return None, None
# Format items as "namespace/name" for display
formatted_items = [f"{ns}/{name}" for name, ns in items]
logging.debug(f"Found {len(formatted_items)} pods")
try:
# Use iterfzf for selection
selected = iterfzf(formatted_items)
if selected:
namespace, name = selected.split("/")
logging.debug(f"Selected pod {name} in namespace {namespace}")
return name, namespace
else:
logging.debug("No selection made")
return None, None
except Exception as e:
logging.error(f"Error during selection: {e}", exc_info=True)
typer.echo(f"Error during selection: {e}", err=True)
raise typer.Exit(1)
def get_pods(namespace=None):
"""Get list of pods in the specified namespace or all namespaces"""
config.load_kube_config()
v1 = client.CoreV1Api()
try:
if namespace:
logging.debug(f"Listing pods in namespace {namespace}")
pod_list = v1.list_namespaced_pod(namespace=namespace)
else:
logging.debug("Listing pods in all namespaces")
pod_list = v1.list_pod_for_all_namespaces()
pods = [(pod.metadata.name, pod.metadata.namespace) for pod in pod_list.items]
logging.debug(f"Found {len(pods)} pods")
return pods
except client.exceptions.ApiException as e:
logging.error(f"Error listing pods: {e}")
typer.echo(f"Error listing pods: {e}", err=True)
raise typer.Exit(1)
def get_pod_spec(pod_name, namespace):
config.load_kube_config()
v1 = client.CoreV1Api()
return v1.read_namespaced_pod(pod_name, namespace)
def get_pod_volumes_and_mounts(pod_spec):
"""Extract all volumes and mounts from a pod spec"""
volume_mounts = []
for container in pod_spec.spec.containers:
if container.volume_mounts:
volume_mounts.extend(container.volume_mounts)
# Filter out None values from volume mounts
volume_mounts = [vm for vm in volume_mounts if format_volume_mount(vm)]
# Get all volumes, including device mounts
volumes = []
if pod_spec.spec.volumes:
for v in pod_spec.spec.volumes:
# Handle device mounts
if v.name in ["cache-volume"]:
volumes.append(
client.V1Volume(
name=v.name,
empty_dir=client.V1EmptyDirVolumeSource(medium="Memory"),
)
)
elif v.name in ["coral-device"]:
volumes.append(
client.V1Volume(
name=v.name,
host_path=client.V1HostPathVolumeSource(
path="/dev/apex_0", type="CharDevice"
),
)
)
elif v.name in ["qsv-device"]:
volumes.append(
client.V1Volume(
name=v.name,
host_path=client.V1HostPathVolumeSource(
path="/dev/dri", type="Directory"
),
)
)
else:
volumes.append(v)
# Filter out None values from volumes
volumes = [v for v in volumes if format_volume(v)]
return volume_mounts, volumes
def get_pod_env_and_secrets(api, namespace, pod_name):
pod = api.read_namespaced_pod(pod_name, namespace)
# Get environment variables from the pod
env_vars = []
for container in pod.spec.containers:
if container.env:
for env in container.env:
env_dict = {"name": env.name}
if env.value:
env_dict["value"] = env.value
elif env.value_from:
if env.value_from.config_map_key_ref:
env_dict["valueFrom"] = {
"configMapKeyRef": {
"name": env.value_from.config_map_key_ref.name,
"key": env.value_from.config_map_key_ref.key,
}
}
elif env.value_from.secret_key_ref:
env_dict["valueFrom"] = {
"secretKeyRef": {
"name": env.value_from.secret_key_ref.name,
"key": env.value_from.secret_key_ref.key,
}
}
elif env.value_from.field_ref:
env_dict["valueFrom"] = {
"fieldRef": {
"fieldPath": env.value_from.field_ref.field_path
}
}
env_vars.append(env_dict)
# Get all volume mounts that are secrets
secret_volumes = []
if pod.spec.volumes:
secret_volumes = [v for v in pod.spec.volumes if v.secret]
return env_vars, secret_volumes
def create_inspector_job(
api, namespace: str, pod_name: str, volume_mounts: list, volumes: list
):
"""Create a Krayt inspector job with the given mounts"""
timestamp = int(time.time())
job_name = f"{pod_name}-krayt-{timestamp}"
# Get environment variables and secrets from the target pod
env_vars, secret_volumes = get_pod_env_and_secrets(api, namespace, pod_name)
# Add secret volumes to our volumes list
volumes.extend(secret_volumes)
# Create corresponding volume mounts for secrets
secret_mounts = []
for vol in secret_volumes:
secret_mounts.append(
{
"name": vol.name,
"mountPath": f"/mnt/secrets/{vol.secret.secret_name}",
"readOnly": True,
}
)
# Convert volume mounts to dictionaries
formatted_mounts = [format_volume_mount(vm) for vm in volume_mounts]
formatted_mounts.extend(secret_mounts)
# Format mount and PVC info for MOTD
mount_info = []
for vm in formatted_mounts:
if vm:
mount_info.append(f"{vm['name']}:{vm['mountPath']}")
pvc_info = []
for v in volumes:
if hasattr(v, "persistent_volume_claim") and v.persistent_volume_claim:
pvc_info.append(f"{v.name}:{v.persistent_volume_claim.claim_name}")
inspector_job = {
"apiVersion": "batch/v1",
"kind": "Job",
"metadata": {
"name": job_name,
"namespace": namespace,
"labels": {"app": "krayt"},
},
"spec": {
"ttlSecondsAfterFinished": 0, # Delete immediately after completion
"template": {
"metadata": {"labels": {"app": "krayt"}},
"spec": {
"containers": [
{
"name": "krayt",
"image": "alpine:latest", # Use Alpine as base for package management
"command": [
"sh",
"-c",
"""
# Install basic tools first
apk update
apk add curl
# Install lf (terminal file manager)
curl -L https://github.com/gokcehan/lf/releases/download/r31/lf-linux-amd64.tar.gz | tar xzf - -C /usr/local/bin
# Install the rest of the tools
apk add ripgrep exa ncdu dust \
file hexyl jq yq bat fd fzf \
htop bottom difftastic \
mtr bind-tools \
aws-cli sqlite sqlite-dev sqlite-libs
# Function to update MOTD
update_motd() {
cat << EOF > /etc/motd
====================================
Krayt Dragon's Lair
====================================
"Inside every volume lies a pearl of wisdom waiting to be discovered."
Mounted Volumes:
$(echo "$MOUNTS" | tr ',' '\\n' | sed 's/^/- /')
Persistent Volume Claims:
$(echo "$PVCS" | tr ',' '\\n' | sed 's/^/- /')
Mounted Secrets:
$(for d in /mnt/secrets/*; do if [ -d "$d" ]; then echo "- $(basename $d)"; fi; done)
Environment Variables:
$(env | sort | sed 's/^/- /')
Your Hunting Tools:
File Navigation:
- lf: Terminal file manager (run 'lf')
- exa: Modern ls (run 'ls', 'll', or 'tree')
- fd: Modern find (run 'fd pattern')
Search & Analysis:
- rg (ripgrep): Fast search (run 'rg pattern')
- bat: Better cat with syntax highlighting
- hexyl: Hex viewer (run 'hexyl file')
- file: File type detection
Disk Usage:
- ncdu: Interactive disk usage analyzer
- dust: Disk usage analyzer
- du: Standard disk usage tool
File Comparison:
- difft: Modern diff tool (alias 'diff')
System Monitoring:
- btm: Modern system monitor (alias 'top')
- htop: Interactive process viewer
JSON/YAML Tools:
- jq: JSON processor
- yq: YAML processor
Network Tools:
- dig: DNS lookup
- mtr: Network diagnostics
Cloud & Database:
- aws: AWS CLI
- sqlite3: SQLite database tool
Type 'tools-help' for detailed usage information
====================================
EOF
}
# Create helpful aliases and functions
cat << 'EOF' > /root/.ashrc
if [ "$PS1" ]; then
cat /etc/motd
fi
# Aliases for better file navigation
alias ls='exa'
alias ll='exa -l'
alias la='exa -la'
alias tree='exa --tree'
alias find='fd'
alias top='btm'
alias diff='difft'
alias cat='bat --paging=never'
# Function to show detailed tool help
tools-help() {
echo "Krayt Dragon Hunter's Guide:"
echo
echo "File Navigation:"
echo " lf : Navigate with arrow keys, q to quit, h for help"
echo " ls, ll, la : List files (exa with different options)"
echo " tree : Show directory structure"
echo " fd pattern : Find files matching pattern"
echo
echo "Search & Analysis:"
echo " rg pattern : Search file contents"
echo " bat file : View file with syntax highlighting"
echo " hexyl file : View file in hex format"
echo " file path : Determine file type"
echo
echo "Disk Usage:"
echo " ncdu : Interactive disk usage analyzer (navigate with arrows)"
echo " dust path : Tree-based disk usage"
echo " du -sh * : Summarize disk usage"
echo
echo "File Comparison:"
echo " diff file1 file2 : Compare files with syntax highlighting"
echo
echo "System Monitoring:"
echo " top (btm) : Modern system monitor"
echo " htop : Process viewer"
echo
echo "JSON/YAML Tools:"
echo " jq . file.json : Format and query JSON"
echo " yq . file.yaml : Format and query YAML"
echo
echo "Network Tools:"
echo " dig domain : DNS lookup"
echo " mtr host : Network diagnostics"
echo
echo "Cloud & Database:"
echo " aws : AWS CLI tool"
echo " sqlite3 : SQLite database tool"
echo
echo "Secrets:"
echo " ls /mnt/secrets : List mounted secrets"
}
# Set some helpful environment variables
export EDITOR=vi
export PAGER=less
EOF
# Set up environment to always source our RC file
echo "export ENV=/root/.ashrc" > /etc/profile
echo "export ENV=/root/.ashrc" > /etc/environment
# Make RC file available to all shells
cp /root/.ashrc /etc/profile.d/motd.sh
ln -sf /root/.ashrc /root/.profile
ln -sf /root/.ashrc /root/.bashrc
ln -sf /root/.ashrc /root/.mkshrc
ln -sf /root/.ashrc /etc/shinit
# Create initial MOTD
update_motd
sleep 3600
""",
],
"env": env_vars
+ [
{"name": "MOUNTS", "value": ",".join(mount_info)},
{"name": "PVCS", "value": ",".join(pvc_info)},
{"name": "ENV", "value": "/root/.ashrc"},
],
"volumeMounts": formatted_mounts,
}
],
"volumes": [format_volume(v) for v in volumes if format_volume(v)],
"restartPolicy": "Never",
},
},
},
}
return inspector_job
PROTECTED_NAMESPACES = {
"kube-system",
"kube-public",
"kube-node-lease",
"argo-events",
"argo-rollouts",
"argo-workflows",
"argocd",
"cert-manager",
"ingress-nginx",
"monitoring",
"prometheus",
"istio-system",
"linkerd",
}
def version_callback(value: bool):
if value:
typer.echo(f"Version: {KRAYT_VERSION}")
raise typer.Exit()
@app.callback(invoke_without_command=True)
def main(
ctx: typer.Context,
version: bool = typer.Option(
False, "--version", "-v", help="Show version", callback=version_callback
),
):
"""
Krack open a Krayt dragon!
"""
if ctx.invoked_subcommand is None:
ctx.get_help()
@app.command()
def exec(
namespace: Optional[str] = typer.Option(
None,
help="Kubernetes namespace. If not specified, will search for inspectors across all namespaces.",
),
):
"""
Enter the Krayt dragon's lair! Connect to a running inspector pod.
If multiple inspectors are found, you'll get to choose which one to explore.
"""
config.load_kube_config()
batch_api = client.BatchV1Api()
try:
if namespace:
logging.debug(f"Listing jobs in namespace {namespace}")
jobs = batch_api.list_namespaced_job(
namespace=namespace, label_selector="app=krayt"
)
else:
logging.debug("Listing jobs in all namespaces")
jobs = batch_api.list_job_for_all_namespaces(label_selector="app=krayt")
running_inspectors = []
for job in jobs.items:
# Get the pod for this job
v1 = client.CoreV1Api()
pods = v1.list_namespaced_pod(
namespace=job.metadata.namespace,
label_selector=f"job-name={job.metadata.name}",
)
for pod in pods.items:
if pod.status.phase == "Running":
running_inspectors.append(
(pod.metadata.name, pod.metadata.namespace)
)
if not running_inspectors:
typer.echo("No running inspector pods found.")
raise typer.Exit(1)
if len(running_inspectors) == 1:
pod_name, pod_namespace = running_inspectors[0]
else:
pod_name, pod_namespace = fuzzy_select(running_inspectors)
if not pod_name:
typer.echo("No inspector selected.")
raise typer.Exit(1)
# Execute the shell
typer.echo(f"Connecting to inspector {pod_namespace}/{pod_name}...")
os.execvp(
"kubectl",
["kubectl", "exec", "-it", "-n", pod_namespace, pod_name, "--", "sh", "-l"],
)
except client.exceptions.ApiException as e:
logging.error(f"Failed to list jobs: {e}")
typer.echo(f"Failed to list jobs: {e}", err=True)
raise typer.Exit(1)
@app.command()
def clean(
namespace: Optional[str] = typer.Option(
None,
help="Kubernetes namespace. If not specified, will cleanup in all namespaces.",
),
yes: bool = typer.Option(
False,
"--yes",
"-y",
help="Skip confirmation prompt.",
),
):
"""
Clean up after your hunt! Remove all Krayt inspector jobs.
Use --yes/-y to skip confirmation and clean up immediately.
"""
config.load_kube_config()
batch_api = client.BatchV1Api()
try:
if namespace:
if namespace in PROTECTED_NAMESPACES:
typer.echo(f"Error: Cannot cleanup in protected namespace {namespace}")
raise typer.Exit(1)
logging.debug(f"Listing jobs in namespace {namespace}")
jobs = batch_api.list_namespaced_job(
namespace=namespace, label_selector="app=krayt"
)
else:
logging.debug("Listing jobs in all namespaces")
jobs = batch_api.list_job_for_all_namespaces(label_selector="app=krayt")
# Filter out jobs in protected namespaces
jobs.items = [
job
for job in jobs.items
if job.metadata.namespace not in PROTECTED_NAMESPACES
]
if not jobs.items:
typer.echo("No Krayt inspector jobs found.")
return
# Show confirmation prompt
if not yes:
job_list = "\n".join(
f" {job.metadata.namespace}/{job.metadata.name}" for job in jobs.items
)
typer.echo(f"The following inspector jobs will be deleted:\n{job_list}")
if not typer.confirm("Are you sure you want to continue?"):
typer.echo("Operation cancelled.")
return
# Delete each job
for job in jobs.items:
try:
logging.debug(
f"Deleting job {job.metadata.namespace}/{job.metadata.name}"
)
batch_api.delete_namespaced_job(
name=job.metadata.name,
namespace=job.metadata.namespace,
body=client.V1DeleteOptions(propagation_policy="Background"),
)
typer.echo(f"Deleted job: {job.metadata.namespace}/{job.metadata.name}")
except client.exceptions.ApiException as e:
logging.error(
f"Failed to delete job {job.metadata.namespace}/{job.metadata.name}: {e}"
)
typer.echo(
f"Failed to delete job {job.metadata.namespace}/{job.metadata.name}: {e}",
err=True,
)
except client.exceptions.ApiException as e:
logging.error(f"Failed to list jobs: {e}")
typer.echo(f"Failed to list jobs: {e}", err=True)
raise typer.Exit(1)
@app.command()
def create(
namespace: Optional[str] = typer.Option(
None,
help="Kubernetes namespace. If not specified, will search for pods across all namespaces.",
),
):
"""
Krack open a Krayt dragon! Create an inspector pod to explore what's inside your volumes.
If namespace is not specified, will search for pods across all namespaces.
The inspector will be created in the same namespace as the selected pod.
"""
pods = get_pods(namespace)
if not pods:
typer.echo("No pods found.")
raise typer.Exit(1)
selected_pod, selected_namespace = fuzzy_select(pods)
if not selected_pod:
typer.echo("No pod selected.")
raise typer.Exit(1)
pod_spec = get_pod_spec(selected_pod, selected_namespace)
volume_mounts, volumes = get_pod_volumes_and_mounts(pod_spec)
inspector_job = create_inspector_job(
client.CoreV1Api(), selected_namespace, selected_pod, volume_mounts, volumes
)
# Output the job manifest
typer.echo(yaml.dump(clean_dict(inspector_job), sort_keys=False))
@app.command()
def version():
"""Show the version of Krayt."""
typer.echo(f"Version: {KRAYT_VERSION}")
if __name__ == "__main__":
app()