[intrepid-notify] corruption in the clusterbank database

Jonathon Anderson janderso at alcf.anl.gov
Thu Aug 14 16:25:20 CDT 2008


Starting 7 August 2008 the procedure for charging allocations for jobs
run on both Surveyor and Intrepid was slightly changed. As part of
this change, an error was made in the calculation of how much of the
allocation was used for a given job that caused jobs to be charged on
a scale of core seconds, rather than the previously used scale of core
minutes. Unfortunately, this error coincided with a parallel effort to
account for a number of jobs which, for a variety of reasons, were not
yet reflected in clusterbank, causing these corrupted charges to be
spread over a wide range of accounting history.

Initial efforts to correct this corruption in-place were unsuccessful.
As an alternative a parallel rebuild of the clusterbank database is
being built, and will become active (replacing the currently active
database) during the regular maintenance period on 18 August 2008.
Until then, data obtained using the cbank utility should be considered
invalid.

I apologize for this inconvenience. If you require further
information, please contact the Argonne LCF support staff at
support at alcf.anl.gov.

~jonathon anderson



More information about the intrepid-notify mailing list