[Gluster-users] very poor performance on Debian Buster
hunter86_bg at yahoo.com
Thu Jun 27 09:52:52 UTC 2019
Let's try to get some info.
1. gluster volume info will provide valuable information
2. Based on previous (step 1) info, you can check the bricks via 'findmnt /path/to/brick/mountpoint' which actually is a mount point for your storage and will hold 1 or more bricks.
/gluster_bricks/volume1/volume1 is my brick and my LV is mounted on /gluster_bricks/volume1
Findmnt will show filesystem, mount point and mount options.
3. Next get some info with iostat to show what is going on the brick (I guess its close to idle, otherwise you won't be here)
4. Check network usage. I prefer iftop, but you can use other stuff
5. Previous steps can show if a brick (or multiple bricks in distributed cluster) is actually a bottleneck of your performance
6. You can get a gluster profile for analysis.
It consists of starting, get info after some time and stop
For details check: https://docs.gluster.org/en/latest/Administrator%20Guide/Monitoring%20Workload/
7. What kind of workload are you uploading ?
Is it miliions of small files, the depth of the directories (dirA/dirB/dirC/dirN ... etc)
8. What is the tuned profile on the gluster nodes ? Use 'tuned-adm active'.
9. What kind of connection do you use - FUSE, libgfapi, built-in nfs/cifs , nfs ganesha ?
Strahil NikolovOn Jun 27, 2019 11:33, richard lucassen <mailinglists at lucassen.org> wrote:
> I run glusterfs server on a sys-V version of Debian Gluster. The
> machine is an 8-core/256GB/SSD server and I want to copy 400GB to a
> mounted gluster device. The copy now runs for more than 3 days and it
> has only copied 243GB. The network activity is around 4 to 8 Mbit.
> Is this a known issue of version 5.5-3? I did not touch the defaults.
> richard lucassen
> Gluster-users mailing list
> Gluster-users at gluster.org
More information about the Gluster-users