Determine statistical likelihood of PG splitting

split_stats.py

#!/usr/bin/env python

import sys
import math
import json as json

debug=False

obj=json.load(sys.stdin)

lowsplit=40*2
highsplit=((40*2)+19)

pgcounter=0
splitcounter={}
couldsplitcount=0

splitthreshold=1.1  # 10 %

for pg in obj['pgmap']['pg_stats']:
    pgcounter+=1
    totfiles=pg['stat_sum']['num_objects']+pg['stat_sum']['num_object_clones']

    couldsplitlevel=0
    possiblelevel=0
    splitlevel=1
    while totfiles>lowsplit*(splitlevel*16):
        possiblelevel=splitlevel
        splitlevel+=1

    if possiblelevel in splitcounter:
        splitcounter[possiblelevel]+=1
    else:
        splitcounter[possiblelevel]=1

    if possiblelevel>0:
        if totfiles<((lowsplit*(possiblelevel*16))*1.1) or totfiles<((highsplit*(possiblelevel*16))*splitthreshold):
            couldsplitlevel=possiblelevel

    if couldsplitlevel>0:
        print pg['pgid']+" within 10% of split level "+str(couldsplitlevel)
        couldsplitcount+=1
    elif(debug):
        print pg['pgid']+" Not within 10% of split level "+str(possiblelevel)+" :: "+str(totfiles)+" / "+str(lowsplit*(possiblelevel*16))+" -> "+str(highsplit*(possiblelevel*16))


print "PG Split statistics:"
for level in splitcounter:
    print "Split level "+str(level)+": "+str(splitcounter[level])+" pgs"

print "\n"+str(couldsplitcount)+" PGs in possible split range of 10%"