Backup tool / Multiple streams / Distributed volume / Spreading

[Florent Delvaille <florent.delvaille@xxxxxxxxx>]

Hello all,

I'm facing a weird issue with GlusterFS used as a Distributed volume to store our backup files. Let me detail it:

We are using Netbackup as a backup solution in our company.

This is made of 1 server managing all backups coming from different clients.

Backups are all done on disk (and then duplicated on tapes).

So, the server is a Linux server, having 1 Gluster share /adv_disk01.

This share, actually a Gluster Distributed volume, is made of 20 bricks of 2TB, coming from our SAN box.

Now, lets come to the issue …

Every policy we start (let say Oracle backup, File backup, NAS backup), seems to be correctly spread between the bricks (3 policies running = 3 bricks 100% used … make sense). So, all good.

But there is a possibility to use multiple streams for the same backup policy.

For example, you could backup 1 Oracle DB, but over 8 streams (so 8 jobs … and basically 8 files). Same for MsSQL and so on.

But, 8 streams (so 8 files), do not actually mean 8 bricks used … while it definitively should be the case.

Here is how the filename is looking:
/SPR-SQL13.MSSQL7.SPR-SQL13.db.DWH_PENSION.~.7.001of008.20150526151807..C

I tried 2 policies, 8 streams each (so, 16 jobs running), only 2 bricks were used …

There is clearly an issue … but what and where …

Hopefully you'll have an idea.

Thx for your feedback.

Regards;

--

Florent Delvaille

_______________________________________________
Gluster-users mailing list
Gluster-users@xxxxxxxxxxx
http://www.gluster.org/mailman/listinfo/gluster-users