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

[Radosław K. <rad...@ko...>]

Hello,

2015-01-23 21:42 GMT+01:00 Dimitri Maziuk <dm...@bm...>:

> 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.
>
>
Yes, I've read. When you do a backup without spooling you can issue any
number of jobs writing to the disk storage as you wish. All of them will be
writing backup stream as fast as it can get it from the client. No any
additional processing is involved. For disk storage it is a fastest way of
doing backups with Bacula. Tested. Sure, you should enable attribute
spooling.

With data spooling you are coping the same data from place to place which
takes time and slow down a backup. Additionally when Bacula despooling the
job stop getting data from a client. Let's count:

- client can supply a backup stream 100MB/s over a 1Gbit/s Ethernet
- server can write or read 100MB/s on the storage
- backup size 6GB (6000MB for rounding in our calculation)
- spooling 10GB - it can spool/despool in one loop

Without spooling we can finish a backup in 60 seconds - 1 minute
With spooling we write 6GB os backup stream to the disk spool area which
take 1 minute, next we need to read spooled data from spooling area and
write it to the main storage which takes another 60 seconds. Overall backup
time is 2 minutes.

Summary:
1 minute for backup job without spooling, backup speed = 100MB/s
2 minutes for backups with spooling, backup speed = 50MB/s

Spooling data for disk based backups will slow down every backup 2 times,
which was to prove.

Do you have a problem with multiple jobs write to the disk storage? I never
had this kind of problem in any of my dozen of deployments.

What you should read from the page you provided:
" (...) It probably does not make a lot of sense to enable data spooling if
you are writing to disk files. (...) "

You should gain 2x speedup in your all backups when you disable data
spooling.

best regards
-- 
Radosław Korzeniewski
rad...@ko...