2002-09-30 13:58:25

by Michael Clark

[permalink] [raw]
Subject: Re: v2.6 vs v3.0

Hi Kevin,

On 09/30/02 21:05, Kevin Corry wrote:
> EVMS is now up-to-date and running on 2.5.39. You can get the latest kernel
> code from CVS (http://sourceforge.net/cvs/?group_id=25076) or Bitkeepr
> (http://evms.bkbits.net/). There will be a new, full release (1.2) coming out
> this week.

Seems you guys are the furthest ahead for a working logical volume manager
in 2.5. Does the EVMS team plan to send patches for 2.5 before the freeze?

It would be great to have EVMS in 2.5 (assuming the community approves of
EVMS going in). Seems to be very non-invasive touching almost no common code.

How far along are you with the clustering support (distributed locking of
cluster metadata and update notification, etc)? This is what i'm really after.

~mc


2002-09-30 16:17:31

by Kevin Corry

[permalink] [raw]
Subject: Re: v2.6 vs v3.0

On Monday 30 September 2002 08:59, Michael Clark wrote:
> Hi Kevin,
>
> On 09/30/02 21:05, Kevin Corry wrote:
> > EVMS is now up-to-date and running on 2.5.39. You can get the latest
> > kernel code from CVS (http://sourceforge.net/cvs/?group_id=25076) or
> > Bitkeepr (http://evms.bkbits.net/). There will be a new, full release
> > (1.2) coming out this week.
>
> Seems you guys are the furthest ahead for a working logical volume manager
> in 2.5. Does the EVMS team plan to send patches for 2.5 before the freeze?

Yes. We may send something in for review this week.

> It would be great to have EVMS in 2.5 (assuming the community approves of
> EVMS going in). Seems to be very non-invasive touching almost no common
> code.
>
> How far along are you with the clustering support (distributed locking of
> cluster metadata and update notification, etc)? This is what i'm really
> after.

Right now we are talking about ways to use EVMS in a fail-over cluster
environment. E.g.: You have four nodes in a cluster each attached to a large
SAN device. EVMS will provide software fencing of the shared storage so each
node in the cluster will have a private portion of the SAN. EVMS will allow
reassigning of storage to other nodes in the cluster in the event of a node
failure. This approach involves the smallest hit to the existing code and
very little extra kernel code.

More general cluster support, with support for fully-shared storage (and all
of the necessary distributed locking and such) will come in 2003. This will
obviously involve more in-depth code changes.

--
Kevin Corry
[email protected]
http://evms.sourceforge.net/