Files
heartbeat/hbd/server/http.py
T
2026-03-29 20:29:33 -04:00

407 lines
14 KiB
Python

"""HTTP server implementation using aiohttp and jinja2."""
import asyncio
import json
import time
import urllib.parse
import os
import logging
from aiohttp import web
import jinja2
logger = logging.getLogger(__name__)
def _render_template(html_str: str, **context) -> str:
tmpl = jinja2.Template(html_str)
return tmpl.render(**context)
async def start(
host: str,
port: int,
config,
hbdclass,
msgs_getter,
log=None,
email=None,
pushmsg=None,
msg_to_websockets=None,
tcss=None,
DEBUG=0,
verbose=False,
get_now=None,
VER="",
threshold_checker=None,
):
"""Start an aiohttp web server and block until cancelled.
This function is intended to be awaited inside the main asyncio event loop.
"""
get_now = get_now or (lambda: time.time())
async def index(request):
res = []
res.append('<!DOCTYPE HTML PUBLIC "-//IETF//DTD HTML 2.0//EN">')
res.append("<html>")
res.append("<head>")
res.append("<title>Heartbeat</title>")
if tcss:
res.append(tcss)
res.append("</head>")
res.append('<body BGCOLOR = "#FFFFFF" LINK = "#008000" VLINK = "#008000">')
res.append(f"<H2>Heartbeat status {VER}</h2>")
res += hbdclass.ubHost.buildhosttable()
res += hbdclass.ubHost.buildmsgtable(msgs_getter())
res.append(
"<p> %s (%s)</p>"
% (
time.strftime("%H:%M:%S", time.localtime(get_now())),
config.get("tz", "CET-1CDT"),
)
)
res.append("</body></html>")
body = "\n".join(res)
return web.Response(text=body, content_type="text/html")
async def api_hosts(request):
lst = [hbdclass.Host.hosts[h].jsons() for h in hbdclass.Host.hosts]
return web.json_response(json.loads("[" + ",".join(lst) + "]"))
async def api_messages(request):
lst = msgs_getter()[-30:]
return web.json_response(lst)
async def cmd(request):
qa = request.rel_url.query
uname = qa.get("h")
ucmd = qa.get("c")
if not ucmd or not uname:
return web.Response(status=400, text="need h= and c= arguments")
if uname not in hbdclass.Host.hosts:
return web.Response(status=400, text=f"h={uname} not found")
hbdclass.Host.hosts[uname].cmds.append(
("CMD", {"cmd": urllib.parse.unquote(ucmd)})
)
return web.Response(text=f"cmd {uname} queued")
async def drop(request):
qa = request.rel_url.query
uname = qa.get("h")
if not uname:
return web.Response(status=400, text="need h= argument")
if uname not in hbdclass.Host.hosts:
return web.Response(status=400, text=f"h={uname} not found")
if log:
log(uname, "dropped")
del hbdclass.Host.hosts[uname]
return web.Response(text="Done")
async def register(request):
qa = request.rel_url.query
uname = qa.get("h")
if not uname:
return web.Response(status=400, text="need h= argument")
if uname not in hbdclass.Host.hosts:
return web.Response(status=400, text=f"h={uname} not found")
ll = hbdclass.Host.hosts[uname].registerDns()
if log:
log(uname, ll)
return web.Response(text=str(ll))
async def update(request):
qa = request.rel_url.query
uname = urllib.parse.unquote(qa.get("h", ""))
ucode = qa.get("c")
if not ucode or not uname:
return web.Response(status=400, text="need h= and c= arguments")
if uname != "All" and uname not in hbdclass.Host.hosts:
return web.Response(status=400, text=f"h={uname} not found")
if uname != "All":
names = [uname]
else:
names = [n for n in hbdclass.Host.hosts if hbdclass.Host.hosts[n].cver >= 2]
out = []
for n in names:
err = None
try:
r = {"csum": None, "code": ucode}
hbdclass.Host.hosts[n].cmds.append(("UPD", r))
except Exception as e:
err = str(e)
out.append(f"update started for {n}: {err if err else 'OK'}")
return web.Response(text="\n".join(out))
async def live(request):
# render template from hbd/templates/live.html using Jinja2
# Resolve templates directory relative to the hbd package
pkg_dir = os.path.dirname(__file__)
templates_dir = config.get("templates_dir", os.path.join(pkg_dir, "templates"))
env = jinja2.Environment(loader=jinja2.FileSystemLoader(templates_dir))
host = config.get("hb_host", "localhost")
extra_scripts = config.get("http_extra_scripts", "")
host = request.host.split(":")[0]
if config.get("wss_port"):
heartbeat_ws_url = f"wss://{host}:{config['wss_port']}/hbd"
else:
heartbeat_ws_url = f"ws://{host}:{config.get('ws_port', 50005)}/hbd"
tmpl = env.get_template("live.html")
body = tmpl.render(
title="Heartbeat",
header="Heartbeat",
request=request,
heartbeat_ws_url=heartbeat_ws_url,
extra_scripts=extra_scripts,
hosts=[
hbdclass.Host.hosts[h].stateinfo() for h in sorted(hbdclass.Host.hosts)
],
messages=msgs_getter()[-30:],
)
return web.Response(text=body, content_type="text/html")
async def static(request):
"""Serve files from the package static directory.
URL form: /static/<path>
"""
p = request.match_info.get("path", "")
logger.debug("static file requested: %s", p)
base = os.path.abspath(os.path.join(os.path.dirname(__file__), "static"))
# normalize and prevent directory traversal
target = os.path.abspath(os.path.normpath(os.path.join(base, p)))
if not target.startswith(base + os.sep) and target != base:
return web.Response(status=403, text="Forbidden")
if not os.path.exists(target) or not os.path.isfile(target):
return web.Response(status=404, text="Not Found")
logger.info("serving static file: %s", target)
return web.FileResponse(path=target)
async def favicon(request):
"""Serve favicon.ico from the package static directory."""
base = os.path.abspath(os.path.join(os.path.dirname(__file__), "static/images"))
target = os.path.join(base, "favicon.ico")
if not os.path.exists(target) or not os.path.isfile(target):
return web.Response(status=404, text="Not Found")
return web.FileResponse(path=target)
# -------------------------------------------------------------------------
# Plugin Data API Endpoints
# -------------------------------------------------------------------------
async def api_host_plugins(request):
"""Get all plugin data for a specific host."""
hostname = request.match_info.get("hostname")
if hostname not in hbdclass.Host.hosts:
return web.json_response(
{"error": f"Host '{hostname}' not found"},
status=404
)
host = hbdclass.Host.hosts[hostname]
# Get plugin data with most recent sample for each plugin
plugins_summary = {}
for plugin_name, samples in host.plugin_data.items():
if samples:
# Get most recent sample
timestamp, data = samples[-1]
plugins_summary[plugin_name] = {
"timestamp": timestamp,
"data": data,
"sample_count": len(samples),
}
return web.json_response({
"hostname": hostname,
"plugins": plugins_summary,
})
async def api_host_plugin_detail(request):
"""Get detailed data for a specific plugin on a host."""
hostname = request.match_info.get("hostname")
plugin_name = request.match_info.get("plugin_name")
if hostname not in hbdclass.Host.hosts:
return web.json_response(
{"error": f"Host '{hostname}' not found"},
status=404
)
host = hbdclass.Host.hosts[hostname]
# Get limit from query parameter
limit = request.rel_url.query.get("limit", "10")
try:
limit = int(limit)
except ValueError:
limit = 10
# Get plugin data
samples = host.get_plugin_data(plugin_name, limit=limit)
if not samples:
return web.json_response(
{"error": f"No data for plugin '{plugin_name}' on host '{hostname}'"},
status=404
)
# Format samples
formatted_samples = [
{
"timestamp": ts,
"data": data,
}
for ts, data in samples
]
return web.json_response({
"hostname": hostname,
"plugin": plugin_name,
"samples": formatted_samples,
"sample_count": len(formatted_samples),
})
async def api_host_alerts(request):
"""Get alert states for a specific host."""
hostname = request.match_info.get("hostname")
if hostname not in hbdclass.Host.hosts:
return web.json_response(
{"error": f"Host '{hostname}' not found"},
status=404
)
host = hbdclass.Host.hosts[hostname]
# Get alert states
alerts = []
for metric_path, alert_state in host.alert_states.items():
alerts.append(alert_state.to_dict())
# Get summary if threshold_checker available
summary = {"ok": 0, "warning": 0, "critical": 0, "unknown": 0}
if threshold_checker:
summary = threshold_checker.get_alert_summary(host.alert_states)
return web.json_response({
"hostname": hostname,
"alerts": alerts,
"summary": summary,
})
async def api_all_alerts(request):
"""Get all active alerts across all hosts."""
all_alerts = []
for hostname, host in hbdclass.Host.hosts.items():
if threshold_checker:
active_alerts = threshold_checker.get_active_alerts(host.alert_states)
else:
# Fallback if no threshold checker
from hbd.server.threshold import AlertLevel
active_alerts = [
state for state in host.alert_states.values()
if state.level != AlertLevel.OK
]
for alert in active_alerts:
alert_dict = alert.to_dict()
alert_dict["hostname"] = hostname
all_alerts.append(alert_dict)
# Sort by level (critical first) then by hostname
level_order = {"CRITICAL": 0, "WARNING": 1, "UNKNOWN": 2, "OK": 3}
all_alerts.sort(
key=lambda a: (level_order.get(a["level"], 99), a["hostname"], a["metric_path"])
)
# Get summary counts
summary = {"critical": 0, "warning": 0, "unknown": 0, "total": len(all_alerts)}
for alert in all_alerts:
level = alert["level"].lower()
if level in summary:
summary[level] += 1
return web.json_response({
"alerts": all_alerts,
"summary": summary,
"host_count": len(hbdclass.Host.hosts),
})
# -------------------------------------------------------------------------
# UI Pages
# -------------------------------------------------------------------------
async def plugins_page(request):
"""Render the plugin metrics visualization page."""
pkg_dir = os.path.dirname(__file__)
templates_dir = config.get("templates_dir", os.path.join(pkg_dir, "templates"))
env = jinja2.Environment(loader=jinja2.FileSystemLoader(templates_dir))
# Collect all hosts with plugin data
hosts_with_plugins = []
for hostname in sorted(hbdclass.Host.hosts.keys()):
host = hbdclass.Host.hosts[hostname]
if host.plugin_data:
hosts_with_plugins.append({
"name": hostname,
"plugins": list(host.plugin_data.keys()),
})
tmpl = env.get_template("plugins.html")
body = tmpl.render(
title="Plugin Metrics - Heartbeat",
header="Plugin Metrics",
hosts=hosts_with_plugins,
)
return web.Response(text=body, content_type="text/html")
async def alerts_page(request):
"""Render the alerts dashboard page."""
pkg_dir = os.path.dirname(__file__)
templates_dir = config.get("templates_dir", os.path.join(pkg_dir, "templates"))
env = jinja2.Environment(loader=jinja2.FileSystemLoader(templates_dir))
tmpl = env.get_template("alerts.html")
body = tmpl.render(
title="Alerts Dashboard - Heartbeat",
header="Alerts Dashboard",
)
return web.Response(text=body, content_type="text/html")
app = web.Application()
app.add_routes(
[
web.get("/", index),
web.get("/api/0/hosts", api_hosts),
web.get("/api/0/messages", api_messages),
web.get("/api/0/hosts/{hostname}/plugins", api_host_plugins),
web.get("/api/0/hosts/{hostname}/plugins/{plugin_name}", api_host_plugin_detail),
web.get("/api/0/hosts/{hostname}/alerts", api_host_alerts),
web.get("/api/0/alerts", api_all_alerts),
web.get("/c", cmd),
web.get("/d", drop),
web.get("/n", register),
web.get("/u", update),
web.get("/live", live),
web.get("/plugins", plugins_page),
web.get("/alerts", alerts_page),
web.get("/static/{path:.*}", static),
web.get("/favicon.ico", favicon),
]
)
runner = web.AppRunner(app)
await runner.setup()
site = web.TCPSite(runner, host, port)
await site.start()
if verbose:
print(f"HTTP server started on {host}:{port}")
try:
await asyncio.Future()
finally:
await runner.cleanup()