[Matplotlib-users] segfault in _backend_agg.so

davidparks21 davidparks21 at gmail.com
Fri Jan 6 20:31:18 EST 2017


I've been using matplotlib with the AGG backend to generate ~10,000 plot
files using PdfPages. I run this over 30 cores using python multiprocessing.
This worked until a recent change.If I run with 30 cores I have a memory
issue after my recent updates (60GB of ram is exhausted and I see
thrashing). If I cut that down to 24 cores I do not run into a memory issue
(~45GB max usage), but I produce the segfault below at exactly 997 plots
generated. I can get to about 1500 plots if I cut that down to 6 cores, and
with 1 core I got that to ~3500 plots before I manually stopped it for lack
of patience.There doesn't seem to be a specific problem with any one plot,
but I encounter the segfault seen below at a seemingly reproducible point
depending on how many cores I run on (the more cores the sooner I encounter
the issue). Memory issues seem like a likely culprit.*I recently added these
constructs to my script:*  inset_axes(...)     # and some plots using the
inset_axes, I use this within subplot  ax.twiny()           # to plot xticks
along the top and bottom*The dmesg error seen in CentOS:*[1387989.122585]
python[24780]: segfault at 7f34a7233bd0 ip 00007f3739fa8ec3 sp
00007ffedcb0aad0 error 6 in _backend_agg.so[7f3739f97000+5a000]I'm also
generating the following warning that seems to be
benign:/soe/davidparks21/anaconda2/lib/python2.7/site-packages/matplotlib/figure.py:1742:
UserWarning: This figure includes Axes that are not compatible with
tight_layout, so its results might be incorrect.  warnings.warn("This figure
includes Axes that are not "



--
View this message in context: http://matplotlib.1069221.n5.nabble.com/segfault-in-backend-agg-so-tp47691.html
Sent from the matplotlib - users mailing list archive at Nabble.com.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mail.python.org/pipermail/matplotlib-users/attachments/20170107/c879bb85/attachment.html>


More information about the Matplotlib-users mailing list