oq engine 2.1.0 ArbitraryMFD error

[Phil Cummins]

Hello,

I have installed OQ Engine version 2.1 on a cluster running Ubuntu 14.04. I upgraded from OQ Engine version 2.0.1 using the instructions in https://github.com/gem/oq-engine/blob/engine-2.1/doc/upgrading/ubuntu.md. The machine has a head node with 3 compute nodes, each of which has 72 CPUs. We are running oq engine on each node separately, with no cross-node communication. 

We did quite a few runs using oq engine, on all the different compute nodes as well as the head node. The results looked fine. One of the source files is attached, which shows each of 13 sources using a single-bin ArbitraryMFD. We ran using this file on one of the compute nodes, and found it failed with the error: 

  File "/usr/lib/python2.7/dist-packages/openquake/commonlib/sourceconvert

er.py", line 216, in split_fault_source_by_magnitude

    min_mag=mag, bin_width=src.mfd.bin_width,

AttributeError: 'ArbitraryMFD' object has no attribute 'bin_width'

This seems odd, since in the OQ manual it says about ArbirtrayMFDs: "There is no bin-width as the rates correspond exactly to the specific magnitude". No bin width is specified in our source model file.

We found at first that the OQ engine on the head node did not produce this error, even though it had been installed in exactly the same way as on the compute nodes. But then after several runs we noticed that it started encountering this error on the head node as well, even though the source input file has not changed.

Fortunately, we also have a development version of OQ engine 2.1.0 installed on an Redhat HPC cluster at my university, and it seemed to run fine there... for a while. Now we find that runs on that machine hit this error as well. With the same source file that ran successfully before. It seems pretty bizarre.

We're also noticing that output is inconsistently written. We use, e.g. "--exports xml,vsv", but no csv gets written. We have "mean_hazard_curves = true" and "hazard_maps = true" in the job.ini file, but we find that mean hazard maps don't get written. This seemed to start happening around the same time we started encountering the ArbirtaryMFD  error (before that I think oq engine always produced the output we asked for), but I don't know if they are related.

Is anyone else experiencing these problems and does anyone have suggestions for work-arounds?

Cheers,

- Phil

[Graeme Weatherill]

Dear Phil,

Yes, sorry this was a bug that was introduced by accident a couple of weeks ago. We fixed it last week (https://github.com/gem/oq-engine/pull/2307). Can you update the version of OpenQuake on your cluster?

Sorry about that!

Graeme

--
You received this message because you are subscribed to the Google Groups "OpenQuake Users" group.
To unsubscribe from this group and stop receiving emails from it, send an email to openquake-use...@googlegroups.com.
For more options, visit https://groups.google.com/d/optout.

[Daniele Viganò]

Dear Phil,

in the couple of days we'll release OQ Engine 2.1.1 which will include the fix Graeme mentioned in the previous email.

Best regards,
Daniele

--

DANIELE VIGANÒ | System Administrator | Skype dennyv85 | +39 0382 5169882

GLOBAL EARTHQUAKE MODEL | working together to assess risk

GEM - globalquakemodel.org | T - @GEMwrld | F - GEMwrld

[Phil Cummins]

Hi Graeme and Daniele,

Thanks for your quick replies. I'll wait for 2.1.1, although our university HPC people may upgrade using the development version.

Best regards,

- Phil

[Michele Simionato]

On Tuesday, October 25, 2016 at 5:25:26 PM UTC+2, Phil Cummins wrote:

We're also noticing that output is inconsistently written. We use, e.g. "--exports xml,vsv", but no csv gets written. We have "mean_hazard_curves = true" and "hazard_maps = true" in the job.ini file, but we find that mean hazard maps don't get written. This seemed to start happening around the same time we started encountering the ArbirtaryMFD  error (before that I think oq engine always produced the output we asked for), but I don't know if they are related.

I am unable to reproduce your issue. I have just run the demo demos/hazard/LogicTreeCase1ClassicalPSHA with the option --exports=xml,csv and it works as intended, with the mean curves and maps exported. There is probably something strange in the files you are using, so please send them to  engine....@openquake.org, if possible.

Incidentally, there is no reason to export in XML, unless you are doing a disaggregation, for which the CSV export is still missing: for simple hazard curves, maps and UHS the CSV export gives the same information of the XML export but it is a lot faster. HTH,

      Michele Simionato