On Wed, Jan 27, 2010 at 2:37 PM, pablort pablort+varn...@gmail.com wrote:
Hello there,
I've sucessfully created graphics based on varnishstat -1 output using cacti
and snmp and I'd really like to do the same thing using varnishtop to graph
TxStatus responses, but it didn't work as I expected.
$ varnishtop -1 -i TxStatus
29481.00 TxStatus
3280.00 TxStatus
1196.00 TxStatus
376.00 TxStatus
23.00 TxStatus
3.00 TxStatus
3.00 TxStatus
The numbers do reflect the TxStatus'es that I see in interactive varnishtop,
but when I try to run it with -1, it doesn't show which entry corresponds to
which status. LOL.
So, is this supposed to be like this or should I file a bug for that ?
Also, what do you guys use for performance analysis ?
I use varnishstat + collectd with a little python module (which means
collectd 4.9.0 or higher, with the python module included -- I can
post the Perl version that the python module replaced, which will work
in pre-4.9.x). I'm only grabbing a few things and aggregating some
(like all LRU events), but it'd be easy to modify to grab other
things.
If you'd like to take a look, I'll paste the module here (but be
forewarned I'm not python wiz, so this good enough for me). In
collectd.conf, along with the regular config, you'll need:
LoadPlugin python
Globals true
/LoadPlugin
Plugin python
ModulePath /path/to/your/collectd/python/modules
Import varnish.stats
# Optional if different from path in module
# Module varnish.stats
#varnishstat_binary /some/other/path/to/varnishstat
# /Module
/Plugin
Here's varnish/stats.py
import collectd, sys, time, subprocess, os.path
from pprint import pformat
varnishstat_binary = /usr/bin/varnishstat
# Map fields
field_map = {
client_req: count_reqs,
cache_hit: count_hits,
n_wrk_failed: count_workerr,
n_wrk_max: count_workerr,
n_wrk_queue: count_workerr,
n_wrk_overflow: count_workerr,
n_wrk_drop: count_workerr,
n_lru_nuked: count_lru,
n_lru_saved: count_lru,
n_lru_moved: count_lru,
}
fields_to_query = ,.join( field_map.keys() )
def varnish_stats_config ( Cfg ):
global varnishstat_binary
for child in Cfg.children:
if child.key == varnishstat_binary:
collectd.debug( [varnish_stats_config] config arg set
key %s: %s
% ( child.key, child.values[0] ) )
varnishstat_binary = child.values[0]
def varnish_stats_init ( ):
if not os.path.exists( varnishstat_binary ):
collectd.error( Can't find varnishstat binary at %s, disabling
plugin % ( varnishstat_binary ) )
collectd.unregister_read( varnish_stats_read )
def varnish_stats_read ( Data=None ):
count_lru, count_hits, count_reqs, count_workerr = 0, 0, 0, 0
stats = {}
fh = subprocess.Popen( [ varnishstat_binary, -1, -f,
fields_to_query ], stdout=subprocess.PIPE )
lines = fh.stdout.readlines()
fh.stdout.close()
for line in lines:
field, val, dummy = line.strip().split( None, 2 )
if field_map.has_key( field ):
if stats.has_key( field_map[ field ] ):
stats[ field_map[ field ] ] += int( val )
else:
stats[ field_map[ field ] ] = int( val )
for field in ( stats.keys() ):
if stats.has_key( field ):
collectd.debug( [varnish_stats_read] Dispatch %s: %d
% ( field,
stats[ field ] ) )
stats_data = collectd.Values( type=operations )
stats_data.plugin = varnish-stats
stats_data.dispatch( values=[ stats[ field ] ],
type_instance=field )
else:
continue
collectd.error( [varnish_stats_read] Unable to
dispatch key: %s %
( field ) )
collectd.register_config( varnish_stats_config )
collectd.register_read( varnish_stats_read )
collectd.register_init( varnish_stats_init )
___
varnish-misc mailing list
varnish-misc@projects.linpro.no
http://projects.linpro.no/mailman/listinfo/varnish-misc
