Idea

Using the vmstat command line utility to quickly determine the root cause of performance problems.



In [1]:

    
%less ../dataset/vmstat_loadtest.log

Data Input

In this version, we use a helper library that I've built to read in data sources into pandas' DataFrame.



In [2]:

    
from ozapfdis.linux import vmstat

stats = vmstat.read_logfile("../dataset/vmstat_loadtest.log")
stats.head()









    Out[2]:







  
    
      
      r
      b
      swpd
      free
      buff
      cache
      si
      so
      bi
      bo
      in
      cs
      us
      sy
      id
      wa
      st
    
    
      UTC
      
      
      
      
      
      
      
      
      
      
      
      
      
      
      
      
      
    
  
  
    
      2019-01-13 16:09:33
      0
      0
      6144
      2720868
      41924
      345548
      0
      0
      29
      691
      588
      541
      2
      4
      93
      1
      0
    
    
      2019-01-13 16:09:34
      0
      0
      6144
      2718404
      42276
      347908
      0
      0
      1421
      116
      1789
      7724
      1
      6
      92
      1
      0
    
    
      2019-01-13 16:09:35
      0
      0
      6144
      2718404
      42276
      347912
      0
      0
      0
      44
      1853
      7724
      3
      3
      94
      0
      0
    
    
      2019-01-13 16:09:36
      0
      0
      6144
      2717652
      42276
      347912
      0
      0
      0
      24
      1778
      7022
      1
      3
      95
      0
      0
    
    
      2019-01-13 16:09:37
      0
      0
      6144
      2717652
      42276
      347912
      0
      0
      0
      20
      1753
      7033
      1
      4
      96
      0
      0

Data Selection



In [3]:

    
cpu_data = stats.iloc[:, -5:]
cpu_data.head()









    Out[3]:







  
    
      
      us
      sy
      id
      wa
      st
    
    
      UTC
      
      
      
      
      
    
  
  
    
      2019-01-13 16:09:33
      2
      4
      93
      1
      0
    
    
      2019-01-13 16:09:34
      1
      6
      92
      1
      0
    
    
      2019-01-13 16:09:35
      3
      3
      94
      0
      0
    
    
      2019-01-13 16:09:36
      1
      3
      95
      0
      0
    
    
      2019-01-13 16:09:37
      1
      4
      96
      0
      0

Visualization



In [4]:

    
%matplotlib inline
cpu_data.plot.area();

	r	b	swpd	free	buff	cache	si	so	bi	bo	in	cs	us	sy	id	wa	st
UTC
2019-01-13 16:09:33	0	0	6144	2720868	41924	345548	0	0	29	691	588	541	2	4	93	1	0
2019-01-13 16:09:34	0	0	6144	2718404	42276	347908	0	0	1421	116	1789	7724	1	6	92	1	0
2019-01-13 16:09:35	0	0	6144	2718404	42276	347912	0	0	0	44	1853	7724	3	3	94	0	0
2019-01-13 16:09:36	0	0	6144	2717652	42276	347912	0	0	0	24	1778	7022	1	3	95	0	0
2019-01-13 16:09:37	0	0	6144	2717652	42276	347912	0	0	0	20	1753	7033	1	4	96	0	0