Zitat von ALyarskiy <bacula-forum AT backupcentral DOT com>:

> Zitat von ALyarskiy <bacula-forum < at > backupcentral.com>:
>
> Zitat von ALyarskiy <bacula-forum < at > backupcentral.com>:
>
> How can i limit writing jobs to 1 without limiting spooling jobs?
>
> I want to have only one job writing on the tape but spool other jobs
> in parallel.
>
> This is the default as far as i understand. The tape is locked
> exclusive from one writer while other jobs which run concurrently will
> spool data as long as the maximum spool size is not reached. Works
> fine here but you need a really fast spool area if you want to feed
> something like LTO-4/5/6 at full speed.
>
> Regards
>
> Andreas
>
>
> It is true for jobs with SpoolData=yes, but i have few jobs that
> have SpoolData=no, so they write data right on the tape.
> I want to limit jobs that have SpoolData=no by 1 concurrent, so only
> 1 job could write data on the tape, and permit the rest jobs with
> SpoolData=yes to spool data on local drive concurrently (5-7 streams).
>
> You might try the "Maximum Concurrent Jobs" directive for the tape
> device but i guess with this you will block the despooling until all
> non spooling jobs have passed. Why do you want to have spooling and
> non-spooling jobs anyway for the same device?
>
> Regards
>
> Andreas
>
> Yes, now i use "Maximum Concurrent Jobs" limitation on device level.
> I want to write large jobs right on tape ( usually > 50 Gb) - there  
> are no network bottlenecks for this clients, and the rest (there are  
> may be networks bottlenecks sometimes) spool to the local drive.
>
> When local spool used backup process have to read from client, write  
> on local drive, then read from it and write the backup to the tape.  
> For large backups it is HUGE overhead. Also there is problem when  
> you have 4-5 large backups in same time - local drive becomes  
> bottleneck (i have 2 Gb network). So it is reasonable to have a  
> possibility to limit concurrent jobs at different levels. For  
> example - 6 Jobs running in same time, 2 of them are large and  
> writing directly to the tape one by one, rest 4 are spooling to  
> local drive and waiting those 2 to complete.

It all depends. Spooling data is nearly no overhead if you use some  
cheap local storage as you should. Plugin some SATA or if you can  
afford SSD, do a RAID0 and spooling will give you the following  
benefits:
- Data on tape is written with the spool size per job so your restore  
will be a lot faster
- You will be able to drive your tape at maximum speed regardles of  
network congestion or clients searching files and therefore prevent  
shoe-shining

Of course it would be nice to have some additional controls in some  
situations, but it looks like no one have written code for this until  
now.

Regards

Andreas



------------------------------------------------------------------------------
Everyone hates slow websites. So do we.
Make your web apps faster with AppDynamics
Download AppDynamics Lite for free today:
http://p.sf.net/sfu/appdyn_d2d_nov
_______________________________________________
Bacula-users mailing list
Bacula-users AT lists.sourceforge DOT net
https://lists.sourceforge.net/lists/listinfo/bacula-users