[Gluster-users] Self Heal Sync Speed after 3.7.11 and small file performance

[ML mail]

Hi,
I am also observing bad performance with small files on a GlusterFS 3.7.11 cluster. For example if I unpack the latest Linux kernel tar file it takes roughly 9 minutes whereas on my laptop it takes 30 seconds.
Maybe there are a some paramters on the GlusterFS side which could help to fine tune the hosting of a lof of (mostly small) files? Does anyone have any recommendations here?
RegardsML
 

    On Friday, May 6, 2016 4:59 PM, Kayra Otaner | BilgiO <kayra.otaner at bilgio.com> wrote:
 

 Hello there,
We've had serious CPU utilization issues with pre 3.7.9 GlusterFS, due to huge number of small files. CPU load was going up to 40-50 (on 32 core system) with nothing else running on it.We've observed multiple other Gluster users reporting similar issues, and waited for bug fix to be released. It seems like SHD freeze issue is resolved with 3.7.10 and now we're upgrading 3.7.11 however now we're observing very slow heal speed between nodes.
I see people are mostly using Gluster for storage backend to OpenStack to house large VM images. I've used Gluster back in 2010, before RH purchase, but my use case had smaller number of files back then.Has anyone else have experience with hosting large number of small files on GlusterFS cluster ?
Thank you  

-- 
Kayra OtanerBilgiO A.Ş. -  SecOps ExpertsPGP KeyID : A945251E | Manager, Enterprise Linux Solutionswww.bilgio.com |  TR +90 (532) 111-7240 x 1001 | US +1 (201) 206-2592
_______________________________________________
Gluster-users mailing list
Gluster-users at gluster.org
http://www.gluster.org/mailman/listinfo/gluster-users

  
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://www.gluster.org/pipermail/gluster-users/attachments/20160506/1ebd454c/attachment.html>