Re: Using ASM for shared linux filesystem storage?

On 30.07.2016 14:19, dkoleary_at_olearycomputers.com wrote:

> The dba team and I (UNIX admin) are being asked to create a shared
> ext4 filesystem across a number of nodes. I believe this is to
> enable us to remove Veritas Cluster Filesystem (VCFS); but, that's
> just a suspicion.

That is odd: they want a filesystem that is intended to live in a single block device to be created as replacement for a cluster filesystem? I mean, there are free and open source cluster filesystems around, why not pick those? I am thinking of

• GlusterFS
• GFS
• CephFS
• Lustre
• maybe even OCFS
• more at [1]

For an ext4 you need a single block device. So to have it distributed you need a distributed block device. Maybe you can pull off something like this with DRBD but as far as I understand the whole block device is mirrored on every node. It seems to me that a technical superior solution is a networked file system because at the filesystem level you have much more knowledge about the data, distribution and failover needs than on the block device level. Higher level, more semantic - it's as easy as that.

> So, the questions: has anyone used ASM to create a shared storage
> filesystem across nodes that are not in any way otherwise related to
> oracle products? I know the dbas create shared filesystems via CRS;
> however, I've not seen ext4 being used. Is that possible/supported/a
> good idea?

I'd think it's not a good idea. Minimum, one would need more information where that requirement to make it ext4 came from to come to a more informed conclusion.

> If anyone has any information, documentation, etc, I'd appreciate
> getting a pointer.

I hope I provided some.

Kind regards

        robert

[1]
https://en.wikipedia.org/wiki/List_of_file_systems#Distributed_file_systems

-- 
remember.guy do |as, often| as.you_can - without end
http://blog.rubybestpractices.com/