Core Minutes 5/7/2013
Core Minutes 5/7/2013ScienceTools: (notes from Jim) Just one item from April 30 concerning updates to pyLikelihood. See Science Tools Development Notes for details.
Pass 7 Reprocessing (Tom G.) Pass7 reprocessing is now caught up through 30 April 2013. (ROOT files only — the FITS files are being held hostage to diffuse model and IRF update and release). It will probably take at least a month.
Pass 8 (Tracy) Made a comparison on a fixed data of sizes and set as summarized in the table below:
| Version | Recon size |
|---|---|
| pass 7 lp43 | 77.8 MB |
| Pass 8 No Tree | 98.7 MB |
| Pass 8 Full Tree | 169.3 MB |
| Pass 8 Compressed | 103.8 MB |
Note that even without the tree at all Pass8 produces an additional 20 MB. Likely causes are
It may be that in some cases doubles may be changed back to double32 without doing any damage; to be investigated.
In Pass7 we should have been writing relations (but didn't); for Pass8 we must. This adds an additional 20 MB [later turned out this number is 2 MB].
Stephan Zimmer's work on compression maybe buys us another 10%. (Heather) Stephan has done some testing of the LZMA compression. He ran into a small stumbling block in that the Data Catalog cannot handle the LZMA compressed files yet and Tony would like some assurance that we need this before supporting it. But evaluation can proceed just on the files themselves, without cataloging, and Stephan has promised to do this.
(Tracy) Plan on producing a preliminary tag soon. He guesses that a serious candidate will be ready in about 3 weeks [however new information keeps coming to light. Stay tuned.]
Core talk:
RM (Tom S.) There is a problem with running RM on rhel6: the workflowCallback program didn't exist for rhel6, hence when they activated the new computers that were rhel6 this step couldn't run there. Jobs weren't progressing when they were being passed to the new machines instead of the old rhel5 machines. It's easy enough to work around: use rhel5 version of program; it runs on rhel6. Longer term solution would be to rebuild the program on rhel6. That could take a while to untangle since first we need a rhel6 version of an external it uses. (Heather) Does this problem go away if we move RM on Linux to Jenkins? (Tom S.) Yes (Heather) and so it might be a better use of your time to work on that move. (Tom S.) Probably so. [Late update: building the program on rhel6 was easier than originally thought. He didn't have to rebuild the external, just fix a couple things preventing the successful build of the program against it. That is now done.]
Various other glitches having to do with the move from glastlnx20 to fermilnx-v03 and the addition of rhel6 boxes into the express queue have been dealt with.
Moving Mood (Joanne) believes we have all the pieces we need to reinstate a world-read mirror of Mood on mysql-node03 but has not heard more from Tony about it. (Tom G.) has had conversations with Tony recently. He is working on coming up with the simplest possible design and most bullet-proof move plan to handle all requirements. [Update: Tony now has a comprehensive plan, but no date yet for the move.]
previous
|
minutes index
|
next
 |
J. Bogart, Last Modified: