Last active
August 29, 2015 14:08
-
-
Save linuxkidd/ab96477d539343eadd0d to your computer and use it in GitHub Desktop.
Output OSD size statistics, total PGs per OSD and per-pool PGs per OSD (primary PG count and secondary PG count)
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
#!/usr/bin/env python | |
import sys | |
import simplejson as json | |
obj=json.load(sys.stdin) | |
# osd_data_object osd.id = [ 'osdSize': [size,used,avail,used%], 'TotalPGs': [primary, secondary], 'PoolName1': [primary, secondary] , 'PoolName2': [primary, secondary], ... ]; | |
osd_data={} | |
# pool_data_object Pool.id = [ pool object from json ]; | |
pool_data={} | |
dfFields=['Size','Used','Avail','Use%'] # Headers for OSD Size data | |
fields=['TotalPGs'] # Begin headers for PG Count data | |
for pool in obj['osdmap']['pools']: # Loop through pools to populate Pool data and Field Headers (with pool_name) | |
pool_data[str(pool['pool'])]=pool # Populate pool_data | |
fields.append(pool['pool_name']) # Populate Field Headers with pool_name | |
for osd in obj['pgmap']['osd_stats']: # Loop through OSD Size statistics | |
osd_data[osd['osd']]={'osdSize': []} # Create osdSize bucket for Size statistics | |
for field in ['kb','kb_used','kb_avail']: # Loop through the fields we care about | |
osd_data[osd['osd']]['osdSize'].append(osd[field]) # Append fields to osdSize bucket | |
if osd['kb'] > 0: | |
osd_data[osd['osd']]['osdSize'].append("{0:0.2f}".format(float(osd['kb_used'])/osd['kb']*100)) # Append Use Percentage formatted to 2 decimal points | |
else: | |
osd_data[osd['osd']]['osdSize'].append("0.00") | |
for field in fields: # Loop through All Fields | |
osd_data[osd['osd']][field]={'PrimPG#': 0, 'SecPG#': 0} # Create PG Count Buckets | |
for pg in obj['pgmap']['pg_stats']: # Loop through ALL PGs | |
poolID,_=pg['pgid'].split('.',1) # Split the Pool ID from the PG.id | |
primSec='PrimPG#' # Default to Primary PG | |
for osdIndex in range(0,len(pg['acting'])): # Loop through all OSDs in the 'acting' state | |
if osdIndex > 0: # If we're not on the first one, it's a secondary copy | |
primSec='SecPG#' | |
osd_data[pg['acting'][osdIndex]]['TotalPGs'][primSec]+=1 # Increment TotalPG counter for either Pri or Sec copy | |
osd_data[pg['acting'][osdIndex]][pool_data[poolID]['pool_name']][primSec]+=1 # Increment PerPool counter for either Pri or Sec copy | |
## Begin output of per OSD data | |
print ',,'+','.join('' for x in dfFields)+',,,'.join(str(x) for x in fields) # Print top row of header | |
print 'osd,'+','.join(x for x in dfFields)+',pri,sec,sum'.join('' for x in fields)+',pri,sec,sum' # Print 2nd row of header | |
for osdid in osd_data: # Loop through all OSDs | |
osdSum=0 # Reset per-OSD Sum | |
outline=str(osdid) # Start output line with OSD.id | |
for field in range(0,len(dfFields)): # Loop through osdSize fields | |
outline+=','+str(osd_data[osdid]['osdSize'][field]) # Append osdSize statistics to the output line | |
for field in fields: # Loop through PG Count Fields | |
both=osd_data[osdid][field]['PrimPG#']+osd_data[osdid][field]['SecPG#'] # Calculate sum of both Pri and Sec PGs | |
outline+=','+str(osd_data[osdid][field]['PrimPG#'])+','+str(osd_data[osdid][field]['SecPG#'])+','+str(both); # Append Pri, Sec and Sum PG counts | |
print outline # Print output line | |
## Start Summary lines | |
sumLine='Sum' | |
avgLine='Avg' | |
minLine='Min' | |
maxLine='Max' | |
for field in range(0,len(dfFields)): # Loop through osdSize fields | |
if field < 3: # If we're on osd Sizes / use KB, adjust it to Gb to not exceed max int size | |
sumLine+=','+str(sum(osd_data[x]['osdSize'][field]/1024/1024 for x in osd_data))+'g' # Append sum of the osdSize field | |
else: # We're on the Use%, so sum doesn't make sense here, leave it empty | |
sumLine+=',' | |
minLine+=','+str(min(osd_data[x]['osdSize'][field] for x in osd_data)) # Append min value to Summary line | |
maxLine+=','+str(max(osd_data[x]['osdSize'][field] for x in osd_data)) # Append max value to Summary line | |
avgLine+=','+str("{0:0.2f}".format(sum(float(osd_data[x]['osdSize'][field]) for x in osd_data)/len(osd_data))) # Append Average value to Summary line | |
for field in fields: # Loop through all PG Count Fields, appending sum, min, max and avg | |
sumLine+=','+str(sum(osd_data[x][field]['PrimPG#'] for x in osd_data))+','+str(sum(osd_data[x][field]['SecPG#'] for x in osd_data))+','+str(sum(osd_data[x][field]['PrimPG#']+osd_data[x][field]['SecPG#'] for x in osd_data)) | |
minLine+=','+str(min(osd_data[x][field]['PrimPG#'] for x in osd_data))+','+str(min(osd_data[x][field]['SecPG#'] for x in osd_data))+','+str(min(osd_data[x][field]['PrimPG#']+osd_data[x][field]['SecPG#'] for x in osd_data)) | |
maxLine+=','+str(max(osd_data[x][field]['PrimPG#'] for x in osd_data))+','+str(max(osd_data[x][field]['SecPG#'] for x in osd_data))+','+str(max(osd_data[x][field]['PrimPG#']+osd_data[x][field]['SecPG#'] for x in osd_data)) | |
avgLine+=','+str("{0:0.2f}".format(sum(float(osd_data[x][field]['PrimPG#']) for x in osd_data)/len(osd_data)))+','+str("{0:0.2f}".format(sum(float(osd_data[x][field]['SecPG#']) for x in osd_data)/len(osd_data)))+','+str("{0:0.2f}".format(sum(float(osd_data[x][field]['PrimPG#']+osd_data[x][field]['SecPG#']) for x in osd_data)/len(osd_data))) | |
## Print Summary lines | |
print sumLine | |
print avgLine | |
print minLine | |
print maxLine |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment