Hi Jason,
> Whenever a server goes offline it takes ages to import last nights cmu's
> and after that you have to change the DNS settings. Depending on how
> many sites you have you could be at it for ages. I would really like to
> find a more practical and efficient way of doing all of this. How does
> this sound?
>
> 2 identical servers, both with the same sites and users set-up but with
> using their own IP address. Rsync all the web & user directories on a
> regular bases. Run a primary DNS server (on a 3rd server) using a mysql
> database and a script that checks each service is running every Xmins,
> when it sees a service down it can change the mysql database setting the
> appropriate records to point at the backup server.
>
> Say HTTP access fails; script detects failure, database gets altered,
> traffic now goes to back-up machine which would be identical to the
> primary.
>
> Can anybody see any obvious hole in this plan, or know of a better way
> to achieve a similar outcome?
> -You would have to set up each new vsite and user twice, bit of a pain
> -I think for simplicity if one service fails the all services should
> fall over so that when the primary is back online you can just Rsync and
> the primary server would be up-to-date
> -I guess Rsync would have to use relative paths
> (/home/sites/www.example.com/web/) instead of (/home/.sites/34/web/) as
> you couldn't guarantee that the site IDs were the same. Is this possible?
> -Mysql databases would also need to be copied very often so not to lose
> new data, is there anyway to get a mysql to replicate changes to a db
> instantly?
>
> Any feedback will be very helpful
You make some pretty good points there.
As is CMU is not a proper tool for disaster recovery - in fact it is a
disaster all by itself. Sure, in the absence of other workable mechanisms one
has to use what he's got. So everyone uses CMU - for migrations as well as
for backups.
Well ... Brian Smith from NuOnce Networks and I are currently working on
something that'll pretty much solve all these issues. In fact all the things
you listed above prompted us to stick our heads together and think of ways to
improve the usability of BlueQuartz.
The name of the product will be Aventurin{e}.
I don't want to spill too many details now, but this is what I can already
say:
- Run multiple instances of CentOS + BlueQuartz (and other Linux
distributions) on the same server with tight resource management for each
virtual server.
- Clustering support with fail over mechanisms. Two identical servers,
connected via 2nd network interface, doing constant RAID-over-ethernet. If
the primary server goes down, the 2ndary inherits all network settings and
switches itself onto the public network, continuing to serve requests where
the primary failed.
- Backups: No more CMU. Backs up the entire server down to the last file and
allows easy restores within a matter of minutes. Runs independently of the
clustering and therefore provides additional means to preserve your valueable
data.
- Migrations of running (virtual) servers almost without downtime.
- Possibility to import existing CentOS + BlueQuartz server (in fact any Linux
distribution) as virtual server.
- Easy to use and intuitive GUI interface.
- As easy to install as CentOS + BlueQuartz.
It'll be a commercial product, available from NuOnce Networks and
Solarspeed.net directly, or from various selected vendors. Sorry, guys, but
after countless weeks of development we can't release this for free.
Current project status: Second full month of limited public beta testing on
servers running real sites and hosting real clients.
Release date ... well, let's just say: "Soon!". Probably a few more weeks.
http://www.aventurin.net/ has some old screen shots of the GUI from back in
early December. Since then quite a few things changed, but it'll be good
enough to give a rough idea.
--
With best regards,
Michael Stauber
www.solarspeed.net