On 6/8/22 11:56, Eric Bollengier via Bacula-users wrote:
Hello,
On 6/8/22 17:47, Bill Arlofski via Bacula-users wrote:
On 6/6/22 07:37, Josh Fisher wrote:
Why does this choice happen only at job start? I have always thought it
should occur at each volume selection, which might alleviate some of
the
issues with concurrent jobs writing to multi-drive autochangers,
particularly when writing to the same pool. Much of the volume
shuffling
between drives is simply because the job's storage device cannot be
reassigned. This would basically be assigning volume and device as an
atomic pair, meaning volume and device would always be assigned from
within a single critical section, probably where the volume selection
code is now. That pairing would persist as long as it was assigned to
one or more jobs.
Hello Josh!
I am not a developer, so can only answer your question with something
along the lines of "this functionality was probably
decided and designed from the beginning, and some pretty major code
changes would be required to make it as flexible as you
are describin
g."
I'd love to hear Eric or another of the developers voice their
opinions from a technical and design point of view here,
including if there are any thoughts or plans to make such changes at
any point in the future.
The feature described by Josh is indeed very exciting, however, the
resource
are allocated/reserved at the job creation, and such dynamic
allocation would
require a good amount of changes. Not impossible, but not part of the
new version.
I don't know if I would be too happy to see my job spread over
multiple places
within a single session, but that's true, if a storage resource
becomes full,
it would be nice to continue elsewhere automatically (and maybe
fill-up all
resources everywhere !)
A new Device+Volume selection would only be made if the volume it was
writing to became full. It might also be possible to handle other i/o
errors by abandoning the error volume and "rewinding" the job to the
point where it began writing to the error volume, then doing another
Device+Volume selection and continuing. It could allow a job to
automatically recover from a bad tape or broken device, etc. assuming
other devices/volumes are available.
It would also still allow multiple jobs to concurrently write to the
same Device+Volume pair, up to the Device's MaximumConcurrentJobs limit.
Thanks for sharing the idea, thanks Bill for the answer!
Best Regards,
Eric
_______________________________________________
Bacula-users mailing list
Bacula-users@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/bacula-users
_______________________________________________
Bacula-users mailing list
Bacula-users@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/bacula-users
