Saturday, May 04, 2019

If it's broke but it doesn't matter

The collaboration week (longer than a week if you include the pre-meetings) is finally over--perhaps now the lustre filesystem will quit wedging. I spent some good sleep time trying to get it back after the evening crashes, and our expert spent a fair bit of morning time on the morning crashes (and my typo). The talks were interesting, and so were the discussions offline--and the rule about experiments being believed by everybody except the experimenter is apt.

One researcher found a problem with how we unpack the waveform from the raw data. It's a real bug, and could conceivably have a detectable impact on low energy analysis and on analysis of events with very many signals. We have been gearing up to re-process all of our data and put the fix into the software running at the Pole.

Except: it doesn't seem to have any impact on any of the analyses! They look the same with and without the fix. We haven't heard from the low energy folks yet--they should have the answer next week. Or maybe later, thanks to the filesystem crashes. I suspect that we won't be implementing the fix anytime soon, and certainly not reprocessing the raw data. Not until the next big bug is found--provided that bug actually makes a difference. Then we'll lump all the fixes together.


FWIW, lustre is free (except for our time), and the rest of the pack is either not quite what we need, less mature, or eye-wateringly expensive.

No comments:

Post a Comment