pybootchartgui.py crash on CentOS 6

[reidrac]

# ./pybootchartgui.py -o out/
No path given, trying /var/log/bootchart.tgz
parsing '/var/log/bootchart.tgz'
parsing 'header'
parsing 'dmesg'
parsing 'cmdline2.log'
parsing 'paternity.log'
parsing 'proc_cpuinfo.log'
parsing 'proc_diskstats.log'
parsing 'proc_meminfo.log'
parsing 'proc_stat.log'
parsing 'taskstats.log'
merged 0 logger processes
pruned 772 process, 0 exploders, 81 threads, and 1 runs
merged 0 logger processes
Traceback (most recent call last):
  File "./pybootchartgui.py", line 23, in <module>
    sys.exit(main())
  File "/root/bootchart/pybootchartgui/main.py", line 161, in main
    render()
  File "/root/bootchart/pybootchartgui/main.py", line 152, in render
    batch.render(writer, res, options, filename)
  File "/root/bootchart/pybootchartgui/batch.py", line 43, in render
    draw.render (ctx, options, 1.0, trace)
  File "/root/bootchart/pybootchartgui/draw.py", line 405, in render
    curr_y = draw_header (ctx, trace.headers, duration)
  File "/root/bootchart/pybootchartgui/draw.py", line 475, in draw_header
    header_y = ctx.font_extents()[2] + 10
MemoryError

[xrmx]

Which version of pybootchartgui? The error means you have run out of memory, this may be a shortage on your machine or maybe an infinite loops or something. Can you please upload /var/log/bootchart.tgz somewhere to help debugging the issue? Thanks!

[reidrac]

I'm using github code and I've tried with both 512MB and 1GB of RAM in the VM.

Finally I tried the same version in a Fedora 14 to process the bootchart.tgz, and it worked. I guess the problem is related to any of the libraries that are shipped with CentOS 6 :(

[xrmx]

Could you please report how much memory pybootchartgui is consuming? The output of this is fine: watch -n 1 "ps aux | grep python" From the data i have here memory grows up to 200MB when drawing the png so i think we have some room for improvement. Need to investigate the issue more.

[reidrac]

I'm afraid I can't report how many memory is using because it crashes too fast.

You can get the bootchart.tgz file here:

http://blackshell.usebox.net/pub/tmp/bootchart.tgz

... although pybootchartgui.py works perfectly with that file in Fedora 14, so I think is related to a problem with any of the CentOS 6 packages.

Another thing, probably unrelated, is that I'm running the OS in runlevel 3 and I'm using EXIT_PROC="mingetty". It doesn't seem to detect mingetty properly because I get the login prompt in about 6 secs but bootchart captures data for about 29 seconds (?).

I hope it helps, I had to setup the VM again :)

[reidrac]

Look, the problem is in a method of ctx, and that comes from:

# from pybootchartgui/batch.py
ctx = cairo.Context (surface)
    draw.render (ctx, options, 1.0, trace)

So I guess it's cairo/pycairo's fault:

pycairo-1.8.6-2.1.el6.x86_64 cairo-1.8.8-3.1.el6.x86_64

If you spot something obvious (ie. cairo misuse), then fix it; but my wild guess is that it's related to these versions of the libraries.

[xrmx]

I cannot reproduce here, pybootchartgui uses at max 50MB with your data, so it may be some bugs in centos 6 old versions of libraries.

wrt mingetty, we wait 20 more seconds before stopping to gather the data when we see the EXIT_PROC in order to get more post-login data for slow systems.

So i'm closing this. Thanks for reporting anyway.

[reidrac]

Thanks for you time.

You may add a note in the documentation regarding the required cairo/pycairo version, saying that you can run pybootchartgui.py in a different system once you got the data :)