----- Original Message -----
> ...
> Would a sensible approach be to flatten (possibly into nothing)
> the result structure before each write in pmlogger? Would you
> expect any oddness with the temporal index with that approach?
> (if so, would an occasional mark record make sense - I imagine
> this introduces some unusual corner cases for pmlogger now -
> such as archives with no data at all!). Any other potential
> gotchas you can foresee here?
Just to followup on a conversation Ken and I had about this;
Ken suggests it should be readily doable with extra pmlogger
logic to reorder valuesets within the pmResult before writing
it out to disk. Moving an empty event metric results to the
end and then decreasing the numpmid field should work.
The other tricky is to ensure the buffer freeing interface sees
the same PDU buffer that was initialised earlier (iow, undoing
the above - just the numpmid bit - before freeing).
cheers.
--
Nathan
|