Web Host Backup Question

Here is my scenario, and I'm hoping that some of you really smart IT guys can help me out here.

My company has a number of web sites that are hosted with a number of different hosting providers. Of course from time-to-time each provider has a problem and a site goes down (still usually within the SLA). What I would like to do ideally is this: with an entirely different host, keep an up-to-date clone of each of these sites so at any given time, I can change the NameServers and will have minimal downtime.

Everything is PHP-based and it hosted on VPS or shared hosting (depending on the site in question). I'm quite sure that everything is on a Linux server using Apache/MySQL (when applicable)/etc.

I have considered doing to following, but not sure if this is sound or not:

1. Create a cron job for each site that creates a gzip backup every night.
2. Also backup the MySQL database if applicable using cron.
3. On the new (backup) server, create cron jobs to retrieve each archive/sql dump and restore them accordingly.

I have also heard of rsyc, but I'm not sure if that's something I can really utilize or not either. If someone with experience with rsyc can chime in here, that would be cool. This tool seems like it could save a lot of time/bandwidth.

Hopefully this makes sense. I appreciate any kind of constructive feedback that I can get. Thank you in advance.

Changing the nameservers and waiting for the DNS to propgate can take a while. It might end up causing more downtime than good.

I assume you don't have any ssl certs on these sites?

rsync would be much better than gzip, and you should be able to use it for the files at least. Mirroring the database is a little harder depending how the hosts configure things.

Do you have shell access to all servers?

http://www.webhostingtalk.com/sho...p?t=524788

Many webhosts provide an automatic failover service that would likely work better.

Quote from LiquidRetro : Changing the nameservers and waiting for the DNS to propgate can take a while. It might end up causing more downtime than good. I assume you don't have any ssl certs on these sites?

Agreed on the downtime. Obviously this would only be switched in rare cases of extreme emergency. I know that propagation can take a while (a few hours). So this would only be done if we knew that our host was going to be down for a lot longer than that. And trust me, it has happened more than once.

As for SSL, we buy those through a third party, so this should not be an issue.

Quote from jkee : Mirroring the database is a little harder depending how the hosts configure things.

Agreed. And a good handful of the sites are wordpress-driven and have databases. I'll have to look into that a bit more.

Quote from jkee : Do you have shell access to all servers?

Most of them, yes. Some of them, no. Some of the sites are hosted with budget hosts like BlueHost, etc.

Quote from jkee : Many webhosts provide an automatic failover service that would likely work better.

I'm just dealing with a number of different hosts with an even greater number of accounts and servers. This could get very expensive. It has even been suggested that I do cross-hosting load balancing, but again, $$$.

Quote from Gordo : Most of them, yes. Some of them, no. Some of the sites are hosted with budget hosts like BlueHost, etc.

I know BlueHost provides SSH access, it's one of features on their website for their most basic of plans.

Quote from menace33 : I know BlueHost provides SSH access, it's one of features on their website for their most basic of plans.

yep, but you used to have to ask for it and fax them a copy of your driver's license -- which quite frankly is ridiculous. They still make you jump through hoops, but at least they aren't asking for photo id anymore.
https://my.bluehost.com/cgi/help/180

Quote from jkee : yep, but you used to have to ask for it and fax them a copy of your driver's license -- which quite frankly is ridiculous. They still make you jump through hoops, but at least they aren't asking for photo id anymore. https://my.bluehost.com/cgi/help/180

No that is a bit ridiculous but if they've found a better and easier authentication method then it is well worth it for the OP.

The only real way to do what you want is a load balancer with a mirror site. Most load balancers use a number of criteria, including if the server is responding, to forward web requests. I find it hard to believe that any "real" hosts have enough downtime to cause you any pain. Is this being hosted out of someone's garage? If you want 100% uptime with a 100% SLA your only real choice is Akamai. They are super expensive. I manage about 20 web sites hosted through them.

Yes, the best way to keep the website static files synchronized is rsync. It is not hard to use. That is what I use to copy the files from the publishing hub to the origin server.

Since you are using PHP (Drupal?) there really is no good way to have redundancy. I am currently evaluating Drupal and from what I see, it is not an Enterprise content management system. More for a small site, personal site, or a site that doesn't require high availability.

Akamai and most content distribution networks only are good for static content. Using an active content management system like Drupal or Wordpress or even Tomcat applications makes it quite a bit harder to build redundancy into your systems. Drupal has a plugin that exports static pages. So depending on what is on the page, you may or may not be able to make it stand-alone.

What you propose for MySQL would not create a real-time usable database copy. Look into how to mirror MySQL.

Quote from dealgate : Look into how to mirror MySQL.

This is surprisingly simple to do

Quote from dealgate : The only real way to do what you want is a load balancer with a mirror site. Most load balancers use a number of criteria, including if the server is responding, to forward web requests. I find it hard to believe that any "real" hosts have enough downtime to cause you any pain. Is this being hosted out of someone's garage?

While I agree with you for the most part, you have to realize a couple of things. First, this is coming from my boss. Within reason, I usually just get the project done regardless. Second, this is really only to be used in extreme cases. We had a host (a good reputable one) that had hard drives crash. I don't know how, but we had a group of sites hosted with them that were down for almost a day. I can promise that in that time frame we not only lost money in wasted advertising, but we also lost thousands in revenue. It's worth it for them in their minds to have the peace of mind that we can "flip a switch" and have some control over our site rather than just sit back and hope that the hosting provider can get their crap together quickly.

Quote from dealgate : Yes, the best way to keep the website static files synchronized is rsync. It is not hard to use. That is what I use to copy the files from the publishing hub to the origin server.

Can you point me to some good documentation that might get me started with this? I have read a few things from Google results, but I haven't found a great example yet.

Quote from dealgate : What you propose for MySQL would not create a real-time usable database copy. Look into how to mirror MySQL.

I will do this, thank you.

Quote from jkee : yep, but you used to have to ask for it and fax them a copy of your driver's license -- which quite frankly is ridiculous. They still make you jump through hoops, but at least they aren't asking for photo id anymore. https://my.bluehost.com/cgi/help/180

That's pretty ridiculous actually. All of this just to get shell access? These guys are actually just down the street from me...maybe I'll just drive down there so I can prove that I'm a real human with a real account. Haha.

Here is an rsync example. This one uses certificates so that no password needs to be entered.

rsync man page - http://linux.die.net/man/1/rsync

#!/bin/bash
#Commands Variables
RSYNCMD='/usr/bin/rsync'
#RSYNCOPTS="-vrlptuHSPO --stats"
#RSYNCOPTS="-vrluHSPO --stats"
RSYNCOPTS="-pvrc --stats"
CHMOD="--chmod=u+rwX,g+rwX,o+rX,o-w"
HOME=`echo $HOME`
# v - verbose, u - update, H - Hard links, S - sparse files
# p - permissions, l - keep symlinks, t - time stamp,
# r - recursive
# a == archive == rlptgoD
# can't use 'a' because it includes goD, group, owner, Devices
RSYNCUSER='wwwweb'
TARGETMACHINE='x.x.x.x'
RSA_FILE="id_rsa-origin-ws-01"
TARGETDIR="/appdata/www/`whoami`"
SOURCE=`hostname`.domain.local
SOURCEDIR="/appdata/www/`whoami`"
LOGFILE="$HOME/bin/`whoami`-rsync.`date '+%Y%m%d-%H%M%S'`"
MAILTO="admin@yourdomain.com"

echo "rsyn log: rsync.`whoami`" > $LOGFILE
echo "rsyncing from $SOURCE $SOURCEDIR to $TARGETMACHINE $TARGETDIR" >> $LOGFILE
echo "" >> $LOGFILE
# Rsynching recursive, verbose, update, Hardlinks, Sparse files, exclude lost+found dir, collect stats, use ssh,
$RSYNCMD $RSYNCOPTS $CHMOD -e "ssh -i $HOME/bin/$RSA_FILE" $SOURCEDIR/ wwwweb@$TARGETMACHINE:$TARGETDIR >> $LOGFILE
#/usr/bin/mailx -s "rsync log: rsync website report `date '+%B %d, %Y'`" $MAILTO < $LOGFILE