A really cool script to write stats from postgres to graphite.

gistfile1.bat

#!/bin/bash
# Copyright 2012 Canonical Ltd.
# Author: JuanJo Ciarlante <jjo@canonical.com>
# License: GPLv2
DATABASE=<your databases separated by spaces>
ROLE=master
: ${PUSH_CMD:="nc <your corbon host> <your carbon port>"}
: ${FREQ:=10min}
for db in ${DATABASES};do
  (
  metric_prefix="<some prefix for graphite>.${ROLE}.${db}"

  stat=pg_stat_user_tables;keypos=3;nodename=relname;
  psql -A ${db} -c "select * from ${stat};" | awk -v freq=${FREQ} -v timestamp="$(date +%s)" -v keypos=$keypos -v nodename=$nodename -v metric_prefix="${metric_prefix}" -vFS='|' '(NR==1){ for(i=0;i<NF;i++) { fieldname[i]=$i } } (NR>1) { for(i=keypos+1;i<NF;i++) { if (gsub(":"," ", $i)) value=mktime($i); else value=$i  ;printf ("%s.%s.%s.%s.%s %d %d\n", metric_prefix, nodename, $keypos, fieldname[i], freq, value, timestamp); }}'
  stat=pg_statio_user_tables;keypos=3;nodename=relname;
  psql -A ${db} -c "select * from ${stat};" | awk -v freq=${FREQ} -v timestamp="$(date +%s)" -v keypos=$keypos -v nodename=$nodename -v metric_prefix="${metric_prefix}" -vFS='|' '(NR==1){ for(i=0;i<NF;i++) { fieldname[i]=$i } } (NR>1) { for(i=keypos+1;i<NF;i++) { if (gsub(":"," ", $i)) value=mktime($i); else value=$i  ;printf ("%s.%s.%s.%s.%s %d %d\n", metric_prefix, nodename, $keypos, fieldname[i], freq, value, timestamp); }}'
  stat=pg_stat_database;keypos=2;nodename=datname;
  psql -A ${db} -c "select * from ${stat};" | awk -v freq=${FREQ} -v timestamp="$(date +%s)" -v keypos=$keypos -v nodename=$nodename -v metric_prefix="${metric_prefix}" -vFS='|' '(NR==1){ for(i=0;i<NF;i++) { fieldname[i]=$i } } (NR>1) { for(i=keypos+1;i<NF;i++) { if (gsub(":"," ", $i)) value=mktime($i); else value=$i  ;printf ("%s.%s.%s.%s.%s %d %d\n", metric_prefix, nodename, $keypos, fieldname[i], freq, value, timestamp); }}' | sed -nr "s/[.]${db}/.total/p"
  ## slony replication delay:
  nodename=repl_delay;
  value="$(psql -At -c "SELECT MAX(EXTRACT(EPOCH FROM st_lag_time)) FROM _sl.sl_status WHERE st_origin = _sl.getlocalnodeid('_sl') AND st_received!=0;" ${db} 2>/dev/null)"
  test -n "${value}" && printf "%s.%s.%s.%s.%s %s %d\n" ${metric_prefix} ${nodename} "slony" "max" ${FREQ} ${value} $(date +%s)
  nodename=process;
  ps -opid= -C postgres| xargs -I@ sed -rn 's/^se.//p' /proc/@/sched 2>/dev/null|awk -v timestamp="$(date +%s)"  -v freq=${FREQ} -v nodename=$nodename -v metric_prefix="${metric_prefix}" -v OFMT=%d '{ acc[$1]+=$3} /exec_start/ { nr_procs++}; END { acc["nr_procs"]=nr_procs;for (i in acc) printf("%s.%s.%s.%s %d %d\n", metric_prefix, nodename, i, freq, acc[i], timestamp);}'
  )| $PUSH_CMD
done

I believe line 9 should read:

for db in ${DATABASE};do