That's right.

However, you could just use a single server.  For example, base it on
the app role, have Scalr automatically connect an EBS store (e.g. to /
mnt/storage) and then make the MySQL data dir a sym link from /var/lib/
mysql -> /mnt/storage/var-lib-mysql (or whever you call the
directory).
Then use the 'sync-to-all' feature of Scalr to rebundle it as a new
role (my-web-mysql or whever).

This will not provide true HA (High Availability), because if the
instance should go down for whatever reason, it could be as long as
10-15mins before Scalr has noticed, started a replacement instance and
mounted the EBS.
Probably also want to use a script to start the mysql server on the
EBSMounted event from the Scalr farm settings (so it doesn't start
before the data dir exists).

In my experience, EC2 instances only go down unexpectedly vary rarely
(probably when the physical hardware it is running on fails, looses
power or Amazon has some problem with their service).  I've only see
it happen once and that was when EC2 was in beta.

Note also that EBS is a 'dumb' block-device, so if an instance
crashes, it could in theory leave the filesystem in a corrupted state
that can't be remounted.  So, use a journalled filesystem like ext3 to
make that extremely unlikely (the default anyway).

Finally, an EBS store can actually fail, though Amazon says their
reliability approximates 10x more reliable that a physical Hard
Drive.  Hence, you probably want to enable the Scalr 'auto snapshot'
feature to automatically take an EBS snapshot (which is stored in S3)
periodically, so that if disaster strikes (meteor hits the Amazon
facility where your instance & EBS reside), you'll still be able to
create a new EBS volume in another availability zone from the last
snapshot (manually).

You will also be able to grow your system (manually) in future, should
the need rise.  Just fire up some Scalr mysql role instances, setup
the master with a copy of your DB and temporarily make it also a slave
to your existing DB - then just switch the app code to point to the
new DB master.  Then just add more app instances and point them all to
the mysql master/slave hosts.

So, with one EC2 instance, if you reserve it, you likely be running
for less than US$40/mo + Scalr fees.

Hope that helps,
-David.

--~--~---------~--~----~------------~-------~--~----~
You received this message because you are subscribed to the Google Groups 
"scalr-discuss" group.
To post to this group, send email to [email protected]
To unsubscribe from this group, send email to 
[email protected]
For more options, visit this group at 
http://groups.google.com/group/scalr-discuss?hl=en
-~----------~----~----~----~------~----~------~--~---

Reply via email to