Expanding to more than 2 servers?

[Josh]

Thanks to M. David for the guide. I have a business need to expand
across more than 2 servers, after reading through your guide, I didn't
see any suggestions/tips on how to accomplish this. Do you have any
pointers or suggestions?

Thanks.

Josh

[M. David Peterson]

Hi Josh,

On Mon, May 19, 2008 at 10:05 PM, Josh <joshua...@gmail.com> wrote:

Thanks to M. David for the guide.

np. :)

 

I have a business need to expand
across more than 2 servers, after reading through your guide, I didn't
see any suggestions/tips on how to accomplish this. Do you have any
pointers or suggestions?

Are you referring to more than two DRBD nodes (e.g. expanding the overall capacity of your redundant drives) or expanding the number of nodes connecting to the DRBD nodes via NFS?

--
/M:D

M. David Peterson
Co-Founder & Chief Architect, 3rd & Urban, LLC
Email: m.d...@3rdandUrban.com | m.d...@amp.fm
Mobile: (206) 999-0588
http://3rdandUrban.com | http://amp.fm | http://www.oreillynet.com/pub/au/2354

[M. David Peterson]

On Mon, May 19, 2008 at 10:05 PM, Josh <joshua...@gmail.com> wrote:

Thanks to M. David for the guide.

BTW... I would consider this more of a starting point than anything else.  My hope is that via a solid group discussion we can begin to find better, faster, more reliable ways to achieve the end goal of "100%" R/W availability to data.

In other words: If you see areas that you feel can be improved, chances are pretty good they both can and should be.  As such, speak up. :)

[Joshua Boyd]

Primarily more than two DRBD nodes, however it is feasible that I'll need multiple nodes connecting to the nodes via NFS at a later point.  The ideal solution would look something like having multiple DRBD nodes in each of the availability zones, and then have one or two clients in each of the zones as well. 

To start though, just having a few more DRBD nodes spread around to increase redundancy and reliability would be good.  I'm not new to dev/admin, but am new to EC2 and distributed/cloud computing.

Josh

[Joshua Boyd]

I've been eyeing glusterfs, but haven't had a chance to play with it yet. The ultimate goal is to be able to fire off a script that can add a node to the storage array, and another script that can add a client :)

Josh

[M. David Peterson]

On Mon, May 19, 2008 at 10:49 PM, Joshua Boyd <joshua...@gmail.com> wrote:

To start though, just having a few more DRBD nodes spread around to increase redundancy and reliability would be good.  I'm not new to dev/admin, but am new to EC2 and distributed/cloud computing.

Oddly enough, I'm somewhat new to admin (though not new to dev.  Started writing code in '83 @ age 11.  Haven't stopped since ;-)), but not new to EC2.  Thinking "in the cloud" so to speak has become my new passion since S3 first hit the scene in early 2006.  This is an area where I can help.  But I have to admit: I develop more and more respect each day of my life for the admin types:  This area of CS is truly an art form.

Will follow-up more to your last response...

[M. David Peterson]

On Mon, May 19, 2008 at 10:50 PM, Joshua Boyd <joshua...@gmail.com> wrote:

I've been eyeing glusterfs, but haven't had a chance to play with it yet. The ultimate goal is to be able to fire off a script that can add a node to the storage array, and another script that can add a client :)

EC2 persistent storage is the ultimate solution.  Wish I could help make it become publicly available sooner rather than later.

In the mean time, I've been working on a couple ideas... Give me 24 hours to put them to "paper".

[Joshua Boyd]

Cool, thanks MD. I was halfway wondering if it wouldn't be a terrible idea to use mdadm to create a RAID array out of the block devices, so as to expand capacity?

Josh

[Joshua Boyd]

Just as a side note, your script just finished executing on my instances, looks like tun0 didn't come up on either of the two clients. Going to have to dig in and see what's happening.

[Joshua Boyd]

Found the problem, this cropped up 4 separate times :)

vtund: option requires an argument -- L

Now I just need to backtrack through the script and figure out what I need to set the vtund options to.

Josh

[M. David Peterson]

Not good... Just heading to bed.  Back in 8.  Sync up then?

[Joshua Boyd]

Sounds good. Looks like only one instance setup /dev/drbd0 too.

[M. David Peterson]

Ugg!  I've got a book outline due at the end of today that I've been heads down on.  I've got another 3-4 hours of work to do and then should be good to go.

Will ping back this thread later this evening...  Sorry for the delay!