This repository has been archived on 2022-10-07. You can view files and clone it, but cannot push or open issues or pull requests.
skynet-webportal/setup-scripts/health-checker.py

352 lines
12 KiB
Python
Raw Normal View History

2020-09-04 14:12:20 +00:00
#!/usr/bin/env python3
import asyncio
import json
import os
import re
import sys
import time
import traceback
2020-09-04 14:12:20 +00:00
from datetime import datetime, timedelta
import discord
import requests
from bot_utils import setup, send_msg
2020-09-04 14:12:20 +00:00
"""
health-checker reads the /health-check endpoint of the portal and dispatches
messages to a Discord channel.
"""
# Get the container name as an argument or use "sia" as default.
CONTAINER_NAME = "sia"
if len(sys.argv) > 2:
CONTAINER_NAME = sys.argv[2]
# Get the number of hours to look back in the logs or use 1 as default.
CHECK_HOURS = 1
if len(sys.argv) > 3:
CHECK_HOURS = int(sys.argv[3])
# Discord messages have a limit on their length set at 2000 bytes. We use
# a lower limit in order to leave some space for additional message text.
DISCORD_MAX_MESSAGE_LENGTH = 1900
2020-09-04 14:12:20 +00:00
GB = 1 << 30 # 1 GiB in bytes
# Free disk space threshold used for notices and shutting down siad.
FREE_DISK_SPACE_THRESHOLD = 50 * GB
FREE_DISK_SPACE_THRESHOLD_CRITICAL = 20 * GB
2020-09-04 14:12:20 +00:00
bot_token = setup()
client = discord.Client()
# exit_after kills the script if it hasn't exited on its own after `delay` seconds
async def exit_after(delay):
await asyncio.sleep(delay)
os._exit(0)
@client.event
async def on_ready():
await run_checks()
asyncio.create_task(exit_after(3))
async def run_checks():
print("Running Skynet portal health checks")
try:
await check_load_average()
await check_disk()
2020-09-29 10:32:45 +00:00
await check_health()
await check_alerts()
await check_portal_size()
except:
2020-09-04 14:12:20 +00:00
trace = traceback.format_exc()
print("[DEBUG] run_checks() failed.")
2020-09-29 10:32:45 +00:00
await send_msg(
client,
"Failed to run the portal health checks!",
file=trace,
force_notify=True,
)
2020-09-04 14:12:20 +00:00
2020-09-29 10:32:45 +00:00
# check_load_average monitors the system load average value and issues a
# warning message if it exceeds 10.
async def check_load_average():
uptime_string = os.popen("uptime").read().strip()
if sys.platform == "Darwin":
pattern = "^.*load averages: \d*\.\d* \d*\.\d* (\d*\.\d*)$"
else:
pattern = "^.*load average: \d*\.\d*, \d*\.\d*, (\d*\.\d*)$"
load_av = re.match(pattern, uptime_string).group(1)
if float(load_av) > 10:
2020-11-20 18:33:07 +00:00
message = "High system load detected in uptime output: {}".format(
uptime_string)
2020-09-29 10:32:45 +00:00
await send_msg(client, message, force_notify=True)
# check_disk checks the amount of free space on the /home partition and issues
2020-09-07 15:59:39 +00:00
# a warning message if it's under FREE_DISK_SPACE_THRESHOLD GB.
async def check_disk():
2020-09-07 15:59:39 +00:00
# We check free disk space in 1024 byte units, so it's easy to convert.
df = os.popen("df --block-size=1024").read().strip()
volumes = {}
# Iterate over the output, ignoring the header line
for line in df.split("\n")[1:]:
fields = list(filter(None, line.split(" ")))
# -1 is "mounted on", 3 is "available space" in KiB which we want in bytes
volumes[fields[-1]] = int(fields[3]) * 1024
# List of mount point, longest to shortest. We'll use that to find the best
# fit for the volume we want to check.
mount_points = sorted(volumes.keys(), key=len, reverse=True)
wd = os.popen("pwd").read().strip()
vol = ""
for mp in mount_points:
if wd.startswith(mp):
vol = mp
break
if vol == "":
2020-09-29 10:32:45 +00:00
message = "Failed to check free disk space! Didn't find a suitable mount point to check."
return await send_msg(client, message, file=df)
2020-11-20 18:33:07 +00:00
# if we've reached a critical free disk space threshold we need to send proper notice
# and shut down sia container so it doesn't get corrupted
if int(volumes[vol]) < FREE_DISK_SPACE_THRESHOLD_CRITICAL:
free_space_gb = "{:.2f}".format(int(volumes[vol]) / GB)
2020-11-20 18:33:07 +00:00
message = "CRITICAL! Very low disk space: {}GiB, **siad stopped**!".format(
free_space_gb)
inspect = os.popen("docker inspect sia").read().strip()
inspect_json = json.loads(inspect)
if inspect_json[0]["State"]["Running"] == True:
2020-11-20 18:33:07 +00:00
# mark portal as unhealthy
os.popen("docker exec health-check cli/disable")
time.sleep(300) # wait 5 minutes to propagate dns changes
os.popen("docker stop sia") # stop sia container
return await send_msg(client, message, force_notify=True)
# if we're reached a free disk space threshold we need to send proper notice
if int(volumes[vol]) < FREE_DISK_SPACE_THRESHOLD:
free_space_gb = "{:.2f}".format(int(volumes[vol]) / GB)
2020-09-29 10:32:45 +00:00
message = "WARNING! Low disk space: {}GiB".format(free_space_gb)
return await send_msg(client, message, force_notify=True)
2020-09-04 14:12:20 +00:00
# check_health checks /health-check endpoint and reports recent issues
async def check_health():
print("\nChecking portal health status...")
2020-09-04 14:12:20 +00:00
try:
2020-09-29 10:32:45 +00:00
res_check = requests.get("http://localhost/health-check", verify=False)
json_check = res_check.json()
json_critical = requests.get(
"http://localhost/health-check/critical", verify=False
).json()
json_verbose = requests.get(
"http://localhost/health-check/verbose", verify=False
).json()
except:
2020-09-04 14:12:20 +00:00
trace = traceback.format_exc()
print("[DEBUG] check_health() failed.")
2020-09-29 10:32:45 +00:00
return await send_msg(
client, "Failed to run the checks!", file=trace, force_notify=True
)
critical_checks_total = 0
critical_checks_failed = 0
verbose_checks_total = 0
verbose_checks_failed = 0
failed_records = []
2020-09-29 10:32:45 +00:00
failed_records_file = None
time_limit = datetime.utcnow() - timedelta(hours=CHECK_HOURS)
for critical in json_critical:
time = datetime.strptime(critical["date"], "%Y-%m-%dT%H:%M:%S.%fZ")
2020-09-04 14:12:20 +00:00
if time < time_limit:
continue
bad = False
2020-09-29 10:32:45 +00:00
for check in critical["checks"]:
critical_checks_total += 1
if check["up"] == False:
critical_checks_failed += 1
bad = True
if bad:
2020-09-29 10:32:45 +00:00
failed_records.append(critical)
for verbose in json_verbose:
time = datetime.strptime(verbose["date"], "%Y-%m-%dT%H:%M:%S.%fZ")
if time < time_limit:
continue
bad = False
for check in verbose["checks"]:
verbose_checks_total += 1
if check["up"] == False:
verbose_checks_failed += 1
bad = True
if bad:
failed_records.append(verbose)
################################################################################
2020-11-20 18:33:07 +00:00
# create a message
2020-09-29 10:32:45 +00:00
################################################################################
message = ""
force_notify = False
if json_check["disabled"]:
message += "__Portal manually disabled!__ "
elif res_check.status_code is not requests.codes["ok"]:
message += "__Portal down!!!__ "
force_notify = True
if critical_checks_failed:
message += "{}/{} CRITICAL checks failed over the last {} hours! ".format(
critical_checks_failed, critical_checks_total, CHECK_HOURS
)
force_notify = True
2020-09-29 10:32:45 +00:00
else:
2020-11-20 18:33:07 +00:00
message += "All {} critical checks passed. ".format(
critical_checks_total)
2020-09-29 10:32:45 +00:00
if verbose_checks_failed:
message += "{}/{} verbose checks failed over the last {} hours! ".format(
verbose_checks_failed, verbose_checks_total, CHECK_HOURS
)
force_notify = True
2020-09-29 10:32:45 +00:00
else:
2020-11-20 18:33:07 +00:00
message += "All {} verbose checks passed. ".format(
verbose_checks_total)
2020-09-29 10:32:45 +00:00
if len(failed_records):
failed_records_file = json.dumps(failed_records, indent=2)
# send a message if we force notification, there is a failures dump or just once daily (heartbeat) on 1 AM
if force_notify or json_check["disabled"] or failed_records_file or datetime.utcnow().hour == 1:
2020-09-29 10:32:45 +00:00
return await send_msg(
client, message, file=failed_records_file, force_notify=force_notify
)
2020-09-04 14:12:20 +00:00
# contains_string is a simple helper to check if a string contains a string.
# This is faster and easier than regex for word comparisons
def contains_string(string_to_check, string_to_find):
2020-11-20 18:33:07 +00:00
return string_to_find in string_to_check
# check_alerts checks the alerts returned from siad's daemon/alerts API
async def check_alerts():
print("\nChecking portal siad alerts...")
# Execute siac alerts and read the response
cmd_string = "docker exec {} siac alerts".format(CONTAINER_NAME)
siac_alert_output = os.popen(cmd_string).read().strip()
# Initialize variables
2020-11-20 18:33:07 +00:00
num_critical_alerts = 0
num_error_alerts = 0
num_warning_alerts = 0
num_siafile_alerts = 0
siafile_alerts = []
# Pattern strings to search for
critical = 'Severity: critical'
error = 'Severity: error'
warning = 'Severity: warning'
health_of = 'has a health of'
siafile_alert_message = "The SiaFile mentioned in the 'Cause' is below 75% redundancy"
2020-11-20 18:33:07 +00:00
# Split the output by line and check for type of alert and siafile alerts
for line in siac_alert_output.split("\n"):
2020-11-20 18:33:07 +00:00
# Check for the type of alert
if contains_string(line, critical):
num_critical_alerts += 1
if contains_string(line, error):
num_error_alerts += 1
if contains_string(line, warning):
num_warning_alerts += 1
# Check for siafile alerts in alerts. This is so that the alert
# severity can change and this doesn't need to be updated
2020-11-20 18:33:07 +00:00
if contains_string(line, siafile_alert_message):
num_siafile_alerts += 1
if contains_string(line, health_of):
siafile_alerts.append(line)
################################################################################
2020-11-20 18:33:07 +00:00
# create a message
################################################################################
message = ""
force_notify = False
if num_critical_alerts > 0:
message += "{} CRITICAL Alerts found! ".format(num_critical_alerts)
force_notify = True
if num_error_alerts > 0:
message += "{} Error Alerts found! ".format(num_error_alerts)
2020-11-20 18:33:07 +00:00
# Subtract out the siafile alerts from the warning alerts since we announce
# them separately
num_warning_alerts -= num_siafile_alerts
message += "{} Warning Alerts found. ".format(num_warning_alerts)
message += "{} SiaFiles with bad health found. ".format(num_siafile_alerts)
# send a message if we force notification, or just once daily (heartbeat)
# on 1 AM
if force_notify or datetime.utcnow().hour == 1:
return await send_msg(
client, message, file=siac_alert_output, force_notify=force_notify
)
# check_portal_size checks the number of files that the portal is managing to
# determine if it is time to rotate it out
async def check_portal_size():
print("\nChecking portal size...")
# Execute siac renter to check the size of the portal
#
# NOTE: we should leave this as always trying to execute the docker command
# against the sia container as this will then fail for maintenance severs
# were we don't care about this check.
cmd_string = "docker exec sia siac renter"
siac_renter_output = os.popen(cmd_string).read().strip()
# Initialize variables
num_files = 0
max_files = 250000
files_text = "Files"
for line in siac_renter_output.split("\n"):
2020-11-20 18:33:07 +00:00
if contains_string(line, files_text):
for el in line.split():
if el.isdigit():
num_files = int(el)
################################################################################
2020-11-20 18:33:07 +00:00
# create a message
################################################################################
message = ""
force_notify = False
if num_files > max_files:
message += "Portal has {} files! Consider rotating! ".format(num_files)
# send notification when above 40% of the limit
force_notify = num_files > max_files * 1.4
2020-11-20 18:33:07 +00:00
else:
message += "Portal has {} files. ".format(num_files)
# send a message if we force notification, or just once daily (heartbeat) on 1 AM
2020-11-24 14:26:51 +00:00
if force_notify or datetime.utcnow().hour == 1:
return await send_msg(
client, message, force_notify=force_notify
)
2020-09-04 14:12:20 +00:00
client.run(bot_token)