+1 on making instance id, job type, retry count, etc to separate columns To convert the blob to a varchar or text, I'm not sure if a comma-separated name-value would be enough though, since it's a map of string to object(any object).
On Wed, Jun 10, 2009 at 6:47 AM, Alex Boisvert<[email protected]> wrote: > +1 we should eliminate the blob column. > > I think we should at least have a column that can be used to correlate the > job with something else, such as an instance id, and possibly a column for > the job type. Both of these should narrow down the number of entries to a > manageable size for querying and then loading all the matches in memory for > inspection. Retry count is generally applicable to all schedules so I'd > make it a column as well. I don't recall if there are other fields that > are generally applicable... and I think we'll still need a job-specific > details column but it could be a string (varchar) with comma-deliminate > name-value pairs instead of a blob. > > We could also get rid of the 'scheduled' and 'transacted' columns which are > no longer of any use, IIRC. > > alex > > On Wed, Jun 10, 2009 at 12:41 AM, Rafal Rusin <[email protected]> wrote: > >> I suggest an improvement for ODE_JOB database table. There's a blob >> column job details, which is hard to maintain. For example, there's no >> way of doing selects on details entries. >> It'd be great to have those details extracted into separate columns >> instead. >> >> What do think about this? >> >> -- >> Rafał Rusin >> http://www.touk.pl >> http://top.touk.pl >> http://www.mimuw.edu.pl/~rrusin <http://www.mimuw.edu.pl/%7Errusin> >> >
