[cctbxbb] Race condition in bootstrap python installation

Wed May 27 10:10:25 PDT 2015

Dear Nigel,

Benchmarking on my desktop machine shows that for the python package the build times are between 98.4s (nproc=4), 101.3s (3), 107.1s (2) and 124.9s (1). So far we had no problems with parallel building any other packages, which is why I would only limit the parallelism of python. We could indeed successively limit it until we no longer observe it, but that does not guarantee race-condition-free compilation for users.

As for dropping the default I think this makes sense. Particularly when building inside virtual machines a 4-fold parallel build can very quickly eat up all the memory and crash. The output provides no hints how the user can fix it.
One possible fix for both cases would be to run with n processors and then rerun failed steps with 1 processor – but this is far from a clean solution.

I think nproc defaults taking into account available cpus/memory together with a fixed nproc=1 specifically for problematic packages such as base/python is the most sensible route.

-Markus

From: cctbxbb-bounces at phenix-online.org [mailto:cctbxbb-bounces at phenix-online.org] On Behalf Of Nigel Moriarty
Sent: 27 May 2015 09:20
To: cctbx mailing list
Subject: Re: [cctbxbb] Race condition in bootstrap python installation

Markus

I have seen the "make" error on occasion here also. It works fine if I force a build so I agree with the race conditions analysis. However, do we have to drop to 1 when 2 or even 3 may have the same end result and not delay the builds as much. Either way we should may the default 1 and then we can use which ever value works best at each site.

Cheers

Nigel

---
Nigel W. Moriarty
Building 64R0246B, Physical Biosciences Division
Lawrence Berkeley National Laboratory
Berkeley, CA 94720-8235
Phone : 510-486-5709     Email : NWMoriarty at LBL.gov<mailto:NWMoriarty at LBL.gov>
Fax   : 510-486-5909       Web  : CCI.LBL.gov<http://CCI.LBL.gov>

On Wed, May 27, 2015 at 8:28 AM, <markus.gerstel at diamond.ac.uk<mailto:markus.gerstel at diamond.ac.uk>> wrote:
We have recently seen intermittent build failures with the bootstrap base installation at Diamond. These all occurred during the compilation of python and were usually of the form

: gcc -pthread -DNDEBUG -g -fwrapv -O3 -Wall -Wstrict-prototypes -Wl,-rpath=\$ORIGIN/../lib Parser/acceler.o Parser/grammar1.o Parser/listnode.o Parser/node.o Parser/parser.o Parser/parsetok.o Parser/bitset.o Parser/metagrammar.o Parser/firstsets.o Parser/grammar.o Parser/pgen.o Objects/obmalloc.o Python/mysnprintf.o Python/pyctype.o Parser/tokenizer_pgen.o Parser/printgrammar.o Parser/pgenmain.o -lpthread -ldl  -lutil -o Parser/pgen
: Parser/node.o: file not recognized: File truncated
: collect2: ld returned 1 exit status
: make[1]: *** [Parser/pgen] Error 1
: make[1]: Leaving directory `/scratch/jenkins_slave/workspace/dials_bootstrap_platforms/compilationtarget/native/label/dials-ws154/build_dials/base_tmp/Python-2.7.8_cci'
: make: *** [Include/graminit.h] Error 2
: make: *** Waiting for unfinished jobs....
Traceback (most recent call last):
  File "modules/cctbx_project/libtbx/auto_build/install_base_packages.py", line 1194, in <module>
    installer(args=sys.argv, log=sys.stdout)
  File "modules/cctbx_project/libtbx/auto_build/install_base_packages.py", line 184, in __init__
    self.build_dependencies(packages=packages)
  File "modules/cctbx_project/libtbx/auto_build/install_base_packages.py", line 571, in build_dependencies
    getattr(self, 'build_%s'%i)()
  File "modules/cctbx_project/libtbx/auto_build/install_base_packages.py", line 622, in build_python
    self.call('make -j %s install'%(self.nproc), log=log, cwd=python_dir)
  File "modules/cctbx_project/libtbx/auto_build/install_base_packages.py", line 264, in call
    return call(args, log=log, verbose=self.verbose, **kwargs)
  File "/scratch/jenkins_slave/workspace/dials_bootstrap_platforms/compilationtarget/native/label/dials-ws154/build_dials/modules/cctbx_project/libtbx/auto_build/installer_utils.py", line 81, in call
    raise RuntimeError("Call to '%s' failed with exit code %d" % (args, rc))
RuntimeError: Call to 'make -j 4 install' failed with exit code 2

Other indicated errors are
  /usr/bin/ld: final link failed: File truncated
or
/usr/bin/ld: can not read symbols: File truncated

I suspect that these errors are manifestations of a race condition in the python build process, which the bootstrap script now by default runs with a -j 4 parallel make.
To test this we are now using --nproc=1 for the base installation step. We haven’t seen these build failures since.
I suggest we set the python compilation to never run with parallel make. Any other ideas?

-Markus

--

This e-mail and any attachments may contain confidential, copyright and or privileged material, and are for the use of the intended addressee only. If you are not the intended addressee or an authorised recipient of the addressee please notify us of receipt by returning the e-mail and do not use, copy, retain, distribute or disclose the information in or attached to the e-mail.
Any opinions expressed within this e-mail are those of the individual and not necessarily of Diamond Light Source Ltd.
Diamond Light Source Ltd. cannot guarantee that this e-mail or any attachments are free from viruses and we cannot accept liability for any damage which you may sustain as a result of software viruses which may be transmitted in or with the message.
Diamond Light Source Limited (company no. 4375679). Registered in England and Wales with its registered office at Diamond House, Harwell Science and Innovation Campus, Didcot, Oxfordshire, OX11 0DE, United Kingdom

_______________________________________________
cctbxbb mailing list
cctbxbb at phenix-online.org<mailto:cctbxbb at phenix-online.org>
http://phenix-online.org/mailman/listinfo/cctbxbb

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://phenix-online.org/pipermail/cctbxbb/attachments/20150527/a799dac9/attachment-0001.htm>