[Halld-online] [New Logentry] Follow-up Re: Corrupted data

[David Lawrence]

Hi All,

  Just as a quick follow up:

I’ve been discussing this today with several members of the DAQ group who also have been 
discussing it among themselves. Graham H., Dave A., Carl T. and I have formed a preliminary
plan for investigation of this and are working closely to understand the problem. I will update
everyone via the e-log and these mailing lists as new information becomes available.

Regards,
-David


> On Jan 6, 2017, at 8:10 AM, davidl at jlab.org wrote:
> 
> Follow-up Re: Corrupted data <https://logbooks.jlab.org/entry/3449814>
> Lognumber 3449814 <https://logbooks.jlab.org/entry/3449814>. Submitted by davidl <https://logbooks.jlab.org/user/davidl> on Fri, 01/06/2017 - 08:07. 
> Logbooks:	HDLOG <https://logbooks.jlab.org/book/hdlog>
> References:	3447847 - Corrupted data <https://logbooks.jlab.org/entry/3447847>
> Attached Files:	
> good_files.txt <https://logbooks.jlab.org/files/2017/01/3449814/good_files.txt>
> bad_files.txt <https://logbooks.jlab.org/files/2017/01/3449814/bad_files.txt>
> bad_runs.txt <https://logbooks.jlab.org/files/2017/01/3449814/bad_runs.txt>
> A larger survey of the files indicates this problem was present throughout the run. Even after the older EVIO 5.1 jar file was reinstated. A total of 425 out of 4410 files indicated some type of error while passing them through hdevio_scan. All corrupted files appeared on /gluonraid3/data3. Roughly half of the runs were written to the data3 partition and the other half to data4. Attached are the lists of "bad" files and "good" ones as well as a list of the runs that contain at least one bad file.
> 
> At this point only one file has been studied in detail which showed the corruption landing on the boundary of a 20 event block. The corruption seemed localized to the start of a single 20 event block so it is possible subsequent data may be recovered with modifications to the parsing code.
> 
> It may be worth noting that some runs (e.g. 22027) had many corrupted files (50/72) but they were not all in a contiguous block. Some of the "good" files in that run happened in between "bad" files.
> 
> _______________________________________________
> Halld-online mailing list
> Halld-online at jlab.org <mailto:Halld-online at jlab.org>
> https://mailman.jlab.org/mailman/listinfo/halld-online <https://mailman.jlab.org/mailman/listinfo/halld-online>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://mailman.jlab.org/pipermail/halld-online/attachments/20170106/7c6ab50c/attachment.html>