Hello All,
İs there any way to implament redistribute already distributed data when new
brick added to DHT distributer ? Means if a new server added and DHT was
used then I would like to balanace capacity and performance with
redistributing data to new added server ?
Regards
Vahric
Maybe you should make some io test with iometer for get it
What kind of raid configurtion will you have on each physcal server ? radi 5 or
raid 6 or raid10 ?
Also whak kind of application will you use , you know DHT is file based and
maybe using it for mysql or mssql couldn't make you happy b
My test servers were not available today, so I had to test my design on
virtual machines :-/
All I can tell is : it works (after correcting minor errors in the
config file). But I have no idea what performances I can expect on real
servers...
Le lundi 01 juin 2009 à 02:35 +0200, Julien Cornuwel a
Hello,
I replacedif (priv->incoming.buf_p)
-free (priv->incoming.buf_p);
With +if (priv->incoming.iobuf)
+iobuf_unref (priv->incoming.iobuf);
And its working .
Now, I will test to see whats going on !
Regards
Vahric
-Original Message-
From
- "Julien Cornuwel" wrote:
> Hi,
>
>
> I will soon have to deploy a glusterFS across several servers. Here
> are
> my goals :
>
> - Every server can access the all filesystem
> - Every file must be replicated
>
> I used a combination of Replicate and Distribute to be able to add
> new
>
Shehjar Tikoo wrote:
Ma-ris Ruskulis wrote:
Seems now, when autoscaling is off, glusterfs running stable, at
least I could not kill it with iozone.
Yes, autoscaling is still under consideration as a feature worth having.
Please avoid using it for the time being.
Thanks
Shehjar
So if this f