Porting loganalyser tools to pandas + jupyter notebooks



In [12]:

    
%matplotlib inline
from loganalyser import vmstat, plot

LINUX_VMSTAT



In [13]:

    
cpu = vmstat.parse('data/supercars-oscounters-13102912.tgz', './vmstat.txt.13102912', 'Europe/Berlin')
cpu.head() # note timestamps are in UTC! (workaround for to_json)









    Out[13]:






  
    
      
      idle
      system
      user
      wait
      timestamp
    
  
  
    
      0
      89
      2
      3
      6
      2013-10-29 11:45:00
    
    
      1
      99
      1
      0
      0
      2013-10-29 11:45:05
    
    
      2
      99
      0
      0
      1
      2013-10-29 11:45:10
    
    
      3
      100
      0
      0
      0
      2013-10-29 11:45:15
    
    
      4
      100
      0
      0
      0
      2013-10-29 11:45:20



In [14]:

    
# find out boundaries to select a timeframe -> use ms since epoch
# in a next version boundaries are provided by ipython widgets!
import datetime
from pytz import timezone
print cpu.ix[100]
# proper way with tz http://www.saltycrane.com/blog/2009/05/converting-time-zones-datetime-objects-python/
ts = timezone('Europe/Berlin').localize(datetime.datetime.strptime('2013-10-29 12:53:20', '%Y-%m-%d %H:%M:%S'))
print int(ts.strftime("%s")) * 1000









    



idle                          99
system                         0
user                           1
wait                           0
timestamp    2013-10-29 11:53:20
Name: 100, dtype: object
1383047600000



In [15]:

    
plot.set_styles(['stacked',])









    Out[15]:



In [16]:

    
plot = reload(plot)
plot.draw_graph('stacked', cpu.to_json(orient='records'),
                {'start': 1383047600000, 'end': 1383048600000, 'yaxis': "CPU utilization [%]"})









    Out[16]:



In [ ]:

	idle	system	user	wait	timestamp
0	89	2	3	6	2013-10-29 11:45:00
1	99	1	0	0	2013-10-29 11:45:05
2	99	0	0	1	2013-10-29 11:45:10
3	100	0	0	0	2013-10-29 11:45:15
4	100	0	0	0	2013-10-29 11:45:20