ganglia varnish gmetric状态监控

实现原理：
运行varnishtat -1会显示varnish的状态数据，其中有uptime，这样直接用shell就可以
生成ganglia所需要的状态数据了。

下面是脚本：

#!/bin/sh

VERSION=1.0

GMETRIC='/usr/bin/gmetric'

# if the GMETRIC program isn't installed, compain
if [ ! -e $GMETRIC ]
then
 
exit 1
fi

# get current statistics
/usr/bin/varnishstat -1 > /tmp/.varnish_current2>&1

if [ ! -f '/tmp/.varnish_last' ]; then
  mv /tmp/.varnish_current/tmp/.varnish_last
  exit 0
fi

awk '
  FILENAME ~ /last/&& NF >= 3 {
   old[$1] = $2
  }

  FILENAME ~ /current/&& NF >= 3 {
   new[$1] = $2
  }

  END {
   interval = new["uptime"] - old["uptime"]

   item="client_conn"
   unit[item] = "connections"
   type[item] = "uint32"
   want[item] = new[item]-old[item]

   item="client_conn_per_second"
   unit[item] = "connections/s"
   type[item] = "uint32"
   want[item] = int(want["client_conn"]/interval)

   item="client_drop"
   unit[item] = "connections"
   type[item] = "uint32"
   want[item] = new[item]-old[item]
   
   item="client_req"
   unit[item] = "requests"
   type[item] = "uint32"
   want[item] = new[item]-old[item]
   
   item="client_req_per_second"
   unit[item] = "requests/s"
   type[item] = "uint32"
   want[item] = int(want["client_req"]/interval)
   
   item="cache_hit"
   unit[item] = "hits"
   type[item] = "uint32"
   want[item] = new[item]-old[item]
   
   item="cache_hit_per_second"
   unit[item] = "hits/s"
   type[item] = "uint32"
   want[item] = int(want["cache_hit"]/interval)
   
   item="cache_miss"
   unit[item] = "miss"
   type[item] = "uint32"
   want[item] = new[item]-old[item]

   item="cache_miss_per_second"
   unit[item] = "miss/s"
   type[item] = "uint32"
   want[item] = int(want["cache_miss"]/interval)

   item="backend_fail"
   unit[item] = "count"
   type[item] = "uint32"
   want[item] = new[item]-old[item]

   item="n_object"
   unit[item] = "objects"
   type[item] = "uint32"
   want[item] = new[item]

   item="n_wrk"
   unit[item] = "threads"
   type[item] = "uint32"
   want[item] = new[item]

   item="n_wrk_drop"
   unit[item]="requets"
   type[item]="uint32"
   want[item]=new[item]-old[item]

   item="n_expired"
   unit[item]="objects"
   type[item]="uint32"
   want[item]=new[item]-old[item]

   item="n_lru_nuked"
   unit[item]="objects"
   type[item]="uint32"
   want[item]=new[item]-old[item]

   item="s_sess"
   unit[item]="sessions"
   type[item]="uint32"
   want[item]=new[item]-old[item]

   item="s_sess_per_second"
   unit[item]="sessions/s"
   type[item]="uint32"
   want[item]=int(want["s_sess"]/interval)

   item="s_pass"
   unit[item]="pass"
   type[item]="uint32"
   want[item]=new[item]-old[item]

   item="s_pass_per_second"
   unit[item]="pass/s"
   type[item]="uint32"
   want[item]=int(want["s_pass"]/interval)

   item="n_purge"
   unit[item]="requets"
   type[item]="uint32"
   want[item]=new[item]-old[item]

   item="backend_retry"
   unit[item]="connections"
   type[item]="uint32"
   want[item]=new[item]-old[item]

   item="accept_fail"
   unit[item]="count"
   type[item]="uint32"
   want[item]=new[item]-old[item]

   item="sm_bfree"
   unit[item]="MB"
   type[item]="uint32"
   want[item]=int(new[item]/(1024*1024))

   for(item in want) {
     printf "%s %d %s %s\n", item, want[item], unit[item],type[item]
   }   

   totallookup = want["cache_hit"]+want["cache_miss"]
   if (totallookup > 0) {
     item="hits_rate"
     unit[item]="hits_rate"
     type[item]="float"
     want[item]=want["cache_hit"]*100/totallookup
     printf "%s %f %s %s\n", item, want[item], unit[item],type[item]
   }
   if (want["client_req"] > 0) {
     item="pass_rate"
     unit[item]="pass_rate"
     type[item]="float"
     want[item]=want["s_pass"]*100/want["client_req"]
     printf "%s %f %s %s\n", item, want[item], unit[item],type[item]
   }

  }
' /tmp/.varnish_last /tmp/.varnish_current | { while read n v u t
do
   $GMETRIC$GMETRIC_ARGS --name="varnish_$n" --value="$v" --type="$t"--units="$u"
done
}

mv /tmp/.varnish_current /tmp/.varnish_last

下面是样图，只选了3张：