PDA

View Full Version : Logging level 'minimal'



avitous
03-28-2006, 03:28 PM
Not sure if this is intentional or, more likely, a minor bug, so I'm asking first in forum...

In Spoon (and in Pan when running from command line) I've noticed logging level 'basic' actually emits quite a bit of information: start and stop for each thread, row count indicators every 5K lines, and so on. Naturally I want to try 'minimal', thinking it would give me start and stop indications with perhaps some summary statistics (like overall rows/sec would be nice, but that's missing...).

It appears 'minimal' logging level causes *nothing* to be logged. No start message, no 'finished OK' message, nothing. I'd expect this with logging level 'none' or 'error', but with 'minimal' I'd expect at least a startup and finished message.

Was this intentional, or would this be a valid bug (or feature request)?

MattCasters
03-28-2006, 11:08 PM
When I thought up the logging I though I would need the minimal, but the question is then of-course: what is minimal?
Perhaps we should have a vote on this?
Please, if you can, do make a tracker for this one, we can add a single line after the transformation has ended: finished OK / finished with errors.

The really minimal is Error of-course: that logs nothing but error lines.
None is for the really confident among us...

Cheers,
Matt

avitous
03-29-2006, 10:54 AM
Done (tracker item #1890).

I'd prefer to have a log message emitted at the start of the run as well, and identification of which transformations/jobs were to be performed. Makes it clearer to establish boundaries when troubleshooting problems, especially if the first responder is looking at the log and doesn't know Kettle. Ending log message should identify completion status.

FWIW we're using Kettle at my employer (CNET) at my behest to implement some simple data transport operations (Sybase -> MySQL) in projects I'm working on; nothing fancy from a data warehousing perspective, but eminently useful nonetheless. This means little components (e.g. a data transport package embedding Kettle runtime) gets deployed to production servers and if problems occur operations engineers need to be able to examine logs to quickly isolate root causes; having some concise (but NONEMPTY) logging establishing start/end of responsible code helps this greatly.

MattCasters
03-29-2006, 11:02 AM
You might be interested to hear that Kettle can also store the logging inside a logging table (CTRL-T, Logging tab).
In development I added a new Tab in Spoon that allows you to browse this logging table for the entries that the currently loaded transformation has generated. That allows you to quickly see which runs failed and get the log besides.

So, we do care about logging and your feedback helps to improve.

Thanks for the tracker, entering these in itself is a lot of work.

;-)

Matt