[Gluster-users] Adding bricks to DHT

Vahriç Muhtaryan vahric at doruk.net.tr
Sun Dec 6 21:30:02 UTC 2009


I guess in list many people live which have more experiance then me 
But I saw that glusterfs do not have such functionalty , its scalable of course, how, if you install 100 servers in the beginning

As for me you should care about two ways ; 
	First use gluster for create parallel stroage cluster
	use server cases which can have 12 , 16 , 24 disk slot and hardware raid.

Because of gluster is file system independent, you can use xfs and can grow parititon how you wany or use lvm for increase capacity.

Hope, developers read and improve functionalty for next versions because if we are talking about scale clustered storage , it should be expandable any time and need backend reallocation mechanisim for keep nodes load equal at the backend.

Regards
Vahric  

-----Original Message-----
From: gluster-users-bounces at gluster.org [mailto:gluster-users-bounces at gluster.org] On Behalf Of Roland Rabben
Sent: Sunday, December 06, 2009 7:29 PM
To: gluster-users at gluster.org
Subject: Re: [Gluster-users] Adding bricks to DHT

My problem is that I can't change directory names without a lot of hassle.
We have a hashed directory structure in our application that we rely on GlusterFS to store.

Is there a way of copying files behind GlusterFS' back and have GlusterFS pick up on the change with a self heal?

It seems that a system designed for scalability should tolerate eh... being scaled...?

Regards
Roland

2009/12/6 Vahriç Muhtaryan <vahric at doruk.net.tr>

> One idea , docs said that you should create new directory. Could you 
> create a new directory after add another server and copy files from 
> old directory to new one, then everything will be distributed.
>
> I know that glusterfs do not have reallocation mechanism for 
> distribute data to new added bricks for achive performance and capcity problem.
>
>
> *Now the question is what about adding a new server?
>
> Currently hash works based on directory level distribution. i.e, a 
> given file's parent directory will have information of how the hash 
> numbers are mapped to subvolumes. So, adding new node doesn't disturb 
> any current setup as the files/dirs present already have its 
> information preserved. Whatever new directory gets created, will start 
> considering new volume for scheduling files.
>
> Regards
> Vahric
>
> -----Original Message-----
> From: gluster-users-bounces at gluster.org [mailto:
> gluster-users-bounces at gluster.org] On Behalf Of Roland Rabben
> Sent: Sunday, December 06, 2009 6:29 PM
> To: gluster-users at gluster.org
> Subject: [Gluster-users] Adding bricks to DHT
>
> Hi
> I have a GlusterFS DHT system that I need to expand in a few days. 
> Reading up on the documentation it seems to me that adding bricks 
> won't solve our problem because of how the hash works.
> I have a fairly static folder structure, but the number of files are 
> growing fast. From what I understand, adding bricks will not allow new 
> files to be stored on the new bricks unless they are stored in new directories.
>
> So my questions are:
> 1. How can I make sure the new bricks are used?
> 2. Is there a way to "rebalance" content over the new and old bricks?
>
> I can't tolerate much downtime on my FS.
>
> I am using GlusterFS 2.06.
>
> Regards
> Roland Rabben
>
>


--
Roland Rabben
Founder & CEO Jotta AS
Cell: +47 90 85 85 39
Phone: +47 21 04 29 00
Email: roland at jotta.no




More information about the Gluster-users mailing list