varnish

#!/usr/bin/env python3
# pylint: disable=invalid-name
# pylint: enable=invalid-name

"""Munin plugin to monitor Varnish status.

=head1 NAME

varnish - monitor Varnish status

=head1 APPLICABLE SYSTEMS

Systems running Varnish cache.

=head1 CONFIGURATION

No configuration is required for this plugin.

=head1 NOTES

This plugin is tested on Varnish 4 but it should also work with v5 and v6.

=head1 AUTHOR

Kim B. Heino <b@bbbs.net>

This is based heavily on varnish4_ plugin from Munin contrib
by Kristian Lyngstol <kristian@bohemians.org> / Redpill Linpro AS.

=head1 LICENSE

GPLv2

=head1 MAGIC MARKERS

 #%# family=auto
 #%# capabilities=autoconf

=cut
"""

import json
import subprocess
import sys


def run_binary(arg):
    """Run binary and return output."""
    try:
        cmd = subprocess.Popen(
            arg, shell=False, close_fds=True, bufsize=-1,
            stdout=subprocess.PIPE, stderr=subprocess.PIPE)
        outdata, dummy_errdata = cmd.communicate()
    except OSError:
        return ''
    return outdata.decode('utf-8', 'ignore')


def get_values():
    """Run varnishstat and parse it's output."""
    output = run_binary(['/usr/bin/varnishstat', '-j'])
    try:
        return json.loads(output)
    except (TypeError, ValueError):
        return {}


def config():
    """Print plugin config."""
    # pylint: disable=too-many-statements

    # backend_traffic
    print('multigraph varnish_backend_traffic')
    print('graph_category webserver')
    print('graph_title Backend traffic')
    print('backend_busy.label Backend conn. too many')
    print('backend_busy.min 0')
    print('backend_busy.type DERIVE')
    print('backend_conn.label Backend conn. success')
    print('backend_conn.min 0')
    print('backend_conn.type DERIVE')
    print('backend_retry.label Backend conn. retry')
    print('backend_retry.min 0')
    print('backend_retry.type DERIVE')
    print('backend_unhealthy.label Backend conn. not attempted')
    print('backend_unhealthy.min 0')
    print('backend_unhealthy.warning :1')
    print('backend_unhealthy.type DERIVE')
    print('backend_recycle.label Backend conn. recycles')
    print('backend_recycle.min 0')
    print('backend_recycle.type DERIVE')
    print('backend_fail.label Backend conn. failures')
    print('backend_fail.min 0')
    print('backend_fail.type DERIVE')
    print('backend_toolate.label Backend conn. was closed')
    print('backend_toolate.min 0')
    print('backend_toolate.type DERIVE')
    print('backend_reuse.label Backend conn. reuses')
    print('backend_reuse.min 0')
    print('backend_reuse.type DERIVE')
    print('backend_req.label Backend requests made')
    print('backend_req.min 0')
    print('backend_req.type DERIVE')

    # bad
    print('multigraph varnish_bad')
    print('graph_category webserver')
    print('graph_title Misbehavior')
    print('sess_drop.label Sessions dropped')
    print('sess_drop.type DERIVE')
    print('threads_failed.label Thread creation failed')
    print('threads_failed.type DERIVE')
    print('threads_destroyed.label Threads destroyed')
    print('threads_destroyed.type DERIVE')
    print('thread_queue_len.label Length of session queue')
    print('thread_queue_len.type GAUGE')
    print('sess_pipe_overflow.label Session pipe overflow')
    print('sess_pipe_overflow.type DERIVE')
    print('esi_warnings.label ESI parse warnings (unlock)')
    print('esi_warnings.type DERIVE')
    print('sess_fail.label Session accept failures')
    print('sess_fail.type DERIVE')
    print('backend_busy.label Backend conn. too many')
    print('backend_busy.type DERIVE')
    print('esi_errors.label ESI parse errors (unlock)')
    print('esi_errors.type DERIVE')
    print('SMF_s0_c_fail.label Allocator failures SMF s0')
    print('SMF_s0_c_fail.type DERIVE')
    print('SMA_Transient_c_fail.label Allocator failures SMA Transient')
    print('SMA_Transient_c_fail.type DERIVE')
    print('losthdr.label HTTP header overflows')
    print('losthdr.type DERIVE')
    print('backend_unhealthy.label Backend conn. not attempted')
    print('backend_unhealthy.type DERIVE')
    print('threads_limited.label Threads hit max')
    print('threads_limited.type DERIVE')
    print('fetch_failed.label Fetch failed (all causes)')
    print('fetch_failed.type DERIVE')

    # expunge
    print('multigraph varnish_expunge')
    print('graph_category webserver')
    print('graph_title Object expunging')
    print('graph_order n_expired n_lru_nuked')
    print('n_lru_nuked.label Number of LRU nuked objects')
    print('n_lru_nuked.min 0')
    print('n_lru_nuked.type DERIVE')
    print('n_expired.label Number of expired objects')
    print('n_expired.min 0')
    print('n_expired.type DERIVE')

    # hit_rate
    print('multigraph varnish_hit_rate')
    print('graph_category webserver')
    print('graph_title Hit rates')
    print('graph_order client_req cache_hit cache_miss cache_hitpass')
    print('graph_scale no')
    print('graph_vlabel %')
    print('graph_args -l 0 -u 100 --rigid')
    print('client_req.label Good client requests received')
    print('client_req.graph off')
    print('client_req.min 0')
    print('client_req.type DERIVE')
    print('cache_miss.label Cache misses')
    print('cache_miss.min 0')
    print('cache_miss.draw STACK')
    print('cache_miss.cdef cache_miss,client_req,/,100,*')
    print('cache_miss.type DERIVE')
    print('cache_hit.label Cache hits')
    print('cache_hit.min 0')
    print('cache_hit.draw AREA')
    print('cache_hit.cdef cache_hit,client_req,/,100,*')
    print('cache_hit.type DERIVE')
    print('cache_hitpass.label Cache hits for pass')
    print('cache_hitpass.min 0')
    print('cache_hitpass.draw STACK')
    print('cache_hitpass.cdef cache_hitpass,client_req,/,100,*')
    print('cache_hitpass.type DERIVE')

    # memory_usage
    print('multigraph varnish_memory_usage')
    print('graph_category webserver')
    print('graph_title Memory usage')
    print('graph_vlabel bytes')
    print('graph_args --base 1024')
    print('SMA_Transient_g_bytes.label Bytes outstanding SMA Transient')
    print('SMA_Transient_g_bytes.type GAUGE')
    print('SMA_Transient_g_space.label Bytes available SMA Transient')
    print('SMA_Transient_g_space.type GAUGE')
    print('SMF_s0_g_bytes.label Bytes outstanding SMF s0')
    print('SMF_s0_g_bytes.type GAUGE')
    print('sms_nbytes.label SMS outstanding bytes')
    print('sms_nbytes.type GAUGE')
    print('sms_balloc.label SMS bytes allocated')
    print('sms_balloc.type GAUGE')
    print('SMF_s0_g_space.label Bytes available SMF s0')
    print('SMF_s0_g_space.type GAUGE')
    print('SMA_Transient_c_bytes.label Bytes allocated SMA Transient')
    print('SMA_Transient_c_bytes.type DERIVE')

    # objects
    print('multigraph varnish_objects')
    print('graph_category webserver')
    print('graph_title Number of objects')
    print('graph_order n_object n_objectcore n_vampireobject n_objecthead')
    print('n_object.label Number of objects')
    print('n_object.type GAUGE')
    print('n_vampireobject.label Number of unresurrected objects')
    print('n_vampireobject.type GAUGE')
    print('n_objectcore.label Number of object cores')
    print('n_objectcore.type GAUGE')
    print('n_objecthead.label Number of object heads')
    print('n_objecthead.info Each object head can have one or more object '
          'attached, typically based on the Vary: header')
    print('n_objecthead.type GAUGE')

    # request_rate
    print('multigraph varnish_request_rate')
    print('graph_category webserver')
    print('graph_title Request rates')
    print('graph_order cache_hit cache_hitpass cache_miss backend_conn '
          'backend_unhealthy client_req client_conn')
    print('client_req.label Good client requests received')
    print('client_req.min 0')
    print('client_req.colour 111111')
    print('client_req.type DERIVE')
    print('s_pipe.label Total pipe sessions seen')
    print('s_pipe.min 0')
    print('s_pipe.colour 1d2bdf')
    print('s_pipe.type DERIVE')
    print('sess_conn.label Sessions accepted')
    print('sess_conn.graph ON')
    print('sess_conn.min 0')
    print('sess_conn.colour 444444')
    print('sess_conn.type DERIVE')
    print('cache_miss.label Cache misses')
    print('cache_miss.min 0')
    print('cache_miss.draw STACK')
    print('cache_miss.colour FF0000')
    print('cache_miss.type DERIVE')
    print('backend_conn.label Backend conn. success')
    print('backend_conn.min 0')
    print('backend_conn.colour 995599')
    print('backend_conn.type DERIVE')
    print('s_pass.label Total pass-ed requests seen')
    print('s_pass.min 0')
    print('s_pass.colour 785d0d')
    print('s_pass.type DERIVE')
    print('backend_unhealthy.label Backend conn. not attempted')
    print('backend_unhealthy.min 0')
    print('backend_unhealthy.colour FF55FF')
    print('backend_unhealthy.type DERIVE')
    print('cache_hitpass.label Cache hits for pass')
    print('cache_hitpass.min 0')
    print('cache_hitpass.draw STACK')
    print('cache_hitpass.colour FFFF00')
    print('cache_hitpass.info Hitpass are cached passes: An entry in the '
          'cache instructing Varnish to pass. Typically achieved after a '
          'pass in vcl_fetch.')
    print('cache_hitpass.type DERIVE')
    print('cache_hit.label Cache hits')
    print('cache_hit.min 0')
    print('cache_hit.draw AREA')
    print('cache_hit.colour 00FF00')
    print('cache_hit.type DERIVE')

    # threads
    print('multigraph varnish_threads')
    print('graph_category webserver')
    print('graph_title Thread status')
    print('threads_created.label Threads created')
    print('threads_created.min 0')
    print('threads_created.type DERIVE')
    print('threads_failed.label Thread creation failed')
    print('threads_failed.min 0')
    print('threads_failed.warning :1')
    print('threads_failed.type DERIVE')
    print('threads_destroyed.label Threads destroyed')
    print('threads_destroyed.min 0')
    print('threads_destroyed.warning :1')
    print('threads_destroyed.type DERIVE')
    print('threads_limited.label Threads hit max')
    print('threads_limited.min 0')
    print('threads_limited.type DERIVE')
    print('threads.label Total number of threads')
    print('threads.min 0')
    print('threads.warning 1:')
    print('threads.type GAUGE')

    # transfer_rates
    print('multigraph varnish_transfer_rates')
    print('graph_category webserver')
    print('graph_title Transfer rates')
    print('graph_order s_resp_bodybytes s_resp_hdrbytes')
    print('graph_vlabel bit/s')
    print('graph_args -l 0')
    print('s_resp_bodybytes.label Body traffic')
    print('s_resp_bodybytes.min 0')
    print('s_resp_bodybytes.draw AREA')
    print('s_resp_bodybytes.cdef s_resp_bodybytes,8,*')
    print('s_resp_bodybytes.type DERIVE')
    print('s_resp_hdrbytes.label Header traffic')
    print('s_resp_hdrbytes.min 0')
    print('s_resp_hdrbytes.draw STACK')
    print('s_resp_hdrbytes.cdef s_resp_hdrbytes,8,*')
    print('s_resp_hdrbytes.info HTTP Header traffic. TCP/IP overhead is not '
          'included.')
    print('s_resp_hdrbytes.type DERIVE')

    # uptime
    print('multigraph varnish_uptime')
    print('graph_category webserver')
    print('graph_title Varnish uptime')
    print('graph_scale no')
    print('graph_vlabel days')
    print('uptime.label Management process uptime')
    print('uptime.cdef uptime,86400,/')
    print('uptime.type GAUGE')


def fetch():
    """Print values."""
    # pylint: disable=too-many-statements
    data = get_values()
    if not data:
        return

    # backend_traffic
    print('multigraph varnish_backend_traffic')
    for key in (
            'backend_busy',
            'backend_conn',
            'backend_retry',
            'backend_unhealthy',
            'backend_recycle',
            'backend_fail',
            'backend_toolate',
            'backend_reuse',
            'backend_req',
    ):
        print('{0}.value {1}'.format(key, data['MAIN.' + key]['value']))

    # bad
    print('multigraph varnish_bad')
    for key in (
            'sess_drop',
            'threads_failed',
            'threads_destroyed',
            'thread_queue_len',
            'sess_pipe_overflow',
            'esi_warnings',
            'sess_fail',
            'backend_busy',
            'esi_errors',
            'losthdr',
            'backend_unhealthy',
            'threads_limited',
            'fetch_failed',
    ):
        print('{0}.value {1}'.format(key, data['MAIN.' + key]['value']))
    print('SMA_Transient_c_fail.value {0}'.format(
        data['SMA.Transient.c_fail']['value']))
    print('SMF_s0_c_fail.value {0}'.format(data['SMF.s0.c_fail']['value']))

    # expunge
    print('multigraph varnish_expunge')
    for key in (
            'n_lru_nuked',
            'n_expired',
    ):
        print('{0}.value {1}'.format(key, data['MAIN.' + key]['value']))

    # hit_rate
    print('multigraph varnish_hit_rate')
    print('client_req.value {0}'.format(
        data['MAIN.cache_hit']['value'] +
        data['MAIN.cache_miss']['value'] +
        data['MAIN.cache_hitpass']['value']))
    for key in (
            'cache_miss',
            'cache_hit',
            'cache_hitpass',
    ):
        print('{0}.value {1}'.format(
            key, data['MAIN.' + key]['value']))

    # memory_usage
    print('multigraph varnish_memory_usage')
    print('SMA_Transient_g_bytes.value {0}'.format(
        data['SMA.Transient.g_bytes']['value']))
    print('SMA_Transient_g_space.value {0}'.format(
        data['SMA.Transient.g_space']['value']))
    print('SMF_s0_g_bytes.value {0}'.format(data['SMF.s0.g_bytes']['value']))
    print('sms_nbytes.value {0}'.format(data['MAIN.sms_nbytes']['value']))
    print('sms_balloc.value {0}'.format(data['MAIN.sms_balloc']['value']))
    print('SMF_s0_g_space.value {0}'.format(data['SMF.s0.g_space']['value']))
    print('SMA_Transient_c_bytes.value {0}'.format(
        data['SMA.Transient.c_bytes']['value']))

    # objects
    print('multigraph varnish_objects')
    for key in (
            'n_object',
            'n_vampireobject',
            'n_objectcore',
            'n_objecthead',
    ):
        print('{0}.value {1}'.format(key, data['MAIN.' + key]['value']))

    # request_rate
    print('multigraph varnish_request_rate')
    for key in (
            'client_req',
            's_pipe',
            'sess_conn',
            'cache_miss',
            'backend_conn',
            's_pass',
            'backend_unhealthy',
            'cache_hitpass',
            'cache_hit',
    ):
        print('{0}.value {1}'.format(key, data['MAIN.' + key]['value']))

    # threads
    print('multigraph varnish_threads')
    for key in (
            'threads_created',
            'threads_failed',
            'threads_destroyed',
            'threads_limited',
            'threads',
    ):
        print('{0}.value {1}'.format(key, data['MAIN.' + key]['value']))

    # transfer_rates
    print('multigraph varnish_transfer_rates')
    for key in (
            's_resp_bodybytes',
            's_resp_hdrbytes',
    ):
        print('{0}.value {1}'.format(key, data['MAIN.' + key]['value']))

    # uptime
    print('multigraph varnish_uptime')
    print('uptime.value {0}'.format(data['MAIN.uptime']['value']))


if __name__ == '__main__':
    if len(sys.argv) > 1 and sys.argv[1] == 'autoconf':
        print('yes' if get_values() else 'no (varnish is not running)')
    elif len(sys.argv) > 1 and sys.argv[1] == 'config':
        config()
    else:
        fetch()