Re: [Bacula-users] how to debug a job

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

Hello,

This is not directed to any particular individual person -- I just chose the last email on this topic to keep the thread.

I have alway thought it would not be useful to use spooling when writing to a disk Volume, and the document heads pretty much in that direction, and I am not really convinced that any given Client will finish faster when spooling, but to really know, one would have to run some tests because the performance analysis is a bit complicated if you have really fast spool disks.

If you do not have fast spooling disks, it seems pretty logical that writing will take longer, because it will need to write to disk, read back from disk, and write to another disk with spooling on, and with spooling off, it simply writes to disk. If you enable Attribute spooling (default in 7.0x.), it will not enter the timing equation since it will despool at the end of the job.

Also, even if the blocks are interleaved for multiple jobs with spooling off, seeking will be quite fast, so it is hard to see that there will be any significant speedup in restores by turning on spooling.

All the above was I have always believed, but in thinking about this more, there is one more item that enters the performance equation, and that is JobMedia records (i.e. the catalog index records into the data). With spooling turned on, there will be roughly (depending on the .conf) one JobMedia per despool, while if spooling is turned off and there are multiple simultaneous jobs writing to the same Volume, there is likely to be on JobMedia record per 65K block written to disk.

Summary: with spooling turned on there will likely be only one JobMedia per despool, and with spooling off there will likely be one per block. JobMedia records are inserted one at a time in the catalog, and if there are a lot of them putting them in and getting them back out, then walking through them in a restore will slow down backup and restore performance. This could turn out to be trivial compared to writing then re-reading a spool file, but if you have a slow catalog, it might be important to consider.

Bottom line: contrary to what I previously thought there *might* be some benefits to turning on spooling for disk Volumes, but only if you have really fast spooling disks (or SSD). This is an interesting topic, and it would be nice to see some real performance measures as often what one expects in performance does not correspond to what computers really deliver.

Best regards,
Kern

On 23.01.2015 21:42, Dimitri Maziuk wrote:
> On 01/23/2015 02:33 PM, Radosław Korzeniewski wrote:
>
>> No, it is not working as you described.
>> Data spooling will read stream from the client, then write it to the
>> temporary disk file, then during despool it is stopping a stream from the
>> client, reads a temporary file and write to destination volume.
>> A standard job will never stop reading a stream and direct write data to
>> the volume.
>
> Read the last bulletpoint under "Other considerations" in
> http://www.bacula.org/5.2.x-manuals/en/main/main/Data_Spooling.html#SECTION003130000000000000000
>
> Then read "10 clients" again.
>
>
>
> ------------------------------------------------------------------------------
> New Year. New Location. New Benefits. New Data Center in Ashburn, VA.
> GigeNET is offering a free month of service with a new server in Ashburn.
> Choose from 2 high performing configs, both with 100TB of bandwidth.
> Higher redundancy.Lower latency.Increased capacity.Completely compliant.
> http://p.sf.net/sfu/gigenet
>
>
> _______________________________________________
> Bacula-users mailing list
> Bacula-users AT lists.sourceforge DOT net
> https://lists.sourceforge.net/lists/listinfo/bacula-users

-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1
Comment: Using GnuPG with Thunderbird - http://www.enigmail.net/

iEYEARECAAYFAlTDs3AACgkQNgfoSvWqwEhJagCguOVJ6qplxgWLt5EJK9Ve2roA
mCYAoPU5osv0/pxZcAeIsSeZWESXGOwh
=oNTT
-----END PGP SIGNATURE-----

------------------------------------------------------------------------------
New Year. New Location. New Benefits. New Data Center in Ashburn, VA.
GigeNET is offering a free month of service with a new server in Ashburn.
Choose from 2 high performing configs, both with 100TB of bandwidth.
Higher redundancy.Lower latency.Increased capacity.Completely compliant.
http://p.sf.net/sfu/gigenet

_______________________________________________
Bacula-users mailing list
Bacula-users AT lists.sourceforge DOT net
https://lists.sourceforge.net/lists/listinfo/bacula-users