Web Hosting Talk : Web Hosting Main Forums : Dedicated Server : 2nd server in case 1 server goes down (redunancy?)

[GP-Brad]

We run a mission critical web-based application. I want to purchase a 2nd dedicated server so that if the 1st server goes offline, the second server will be online without anyone seeing downtime, of course they would have to relog in to get new php sessions and cookies.

I'm thinking of doing round-robin dns and then sync my files and databases. but i was told round robin is load balanced and I dont want load balanced because i want everyone on our main server UNTIL it goes offline.

Is this the best way to do it?

Thanks for any help.

[Blapto]

Round-robin isn't seamless.

Really, perfect load balancing isn't possible, you can just get very very close.
Have a look at linux-ha, heartbeat and other high availability projects.

[glued2]

Ye, echo what Blapto said. If you are using Linux (and I assume you are as you mentioned rsync), HA maybe your best solution (HA is high availability). Your DC's setup would need to be such that it would work, and both servers would probably need to be on the same subnet. You'd have 3 IP's one for each server that never changes and a site IP.... heartbeat would run checking the connection between the two boxes, if heartbeat on the backup server can not contact the primary server, it assigns the backup server the site IP - So in a normal network all traffic to the site IP would just go to the only host that's up (at that point the backup server).

It's not without problems:
If loads of data is changing you may want a dedicated 100mb port (with crossover) between the two boxes.
There is also the problem when they fail by accident, (someone pulls a cable out of either machine, someone restarts the network on either machine, someone reboots either machine - without first stopping heartbeat on the backup)
Finally the way heartbeat works this would protect against hardware failure but wouldn't really detect software failure. If apache crashed on the primary server... it wouldn't be detected.

[Alex]

Why not use round-robin? Then you can offset the load on the main server, and be redundant.

[augustino]

Somebody know about of global load balancing?

This, Could it be good for get a less latency?

[Eleven2 Hosting]

Check out lamp/cluster from emicnetworks.com

This is a very solid solution!

[efarmer]

GP-Brad

What is your concern?
Provider problem or hardware problem?

[Alex]

Originally posted by efarmer
GP-Brad

What is your concern?
Provider problem or hardware problem?

I don't think he has a specific conern, I believe he just wants to be redundant. There are large numbers of reasons why his server could go offline.

[GP-Brad]

Originally posted by efarmer
GP-Brad

What is your concern?
Provider problem or hardware problem?

My major concern is network outtages. Im not concerned with hardware.

Here is the scenario

I have 1 linux server currently running CentOS and cpanel. The site I need redundant is about 20mb in size, the database about 10mb in size.

Because of network outtages I want another linux server at a seperate datacenter. This way if the datacenter has problems the 2nd server won't be affected since its in a different location.

so basically i want to send 20mb of files and 1 10mb database over the net like every 20 minutes so the 2nd server is up to date. then if server 1 goes offline, server 2 will be up to date (or within 20 minutes to date)

Thanks

[efarmer]

Since your concern is network outages,
you may have a 2nd server at another location.

Using services like DNS failover by http://www.dnsmadeeasy.com, visitors will be sent over to the 2nd server when your 1st server did not response and then send visitors back to 1st server when it recovered.

Rsync may be setup to duplicate data over from 1st server to 2nd server at your required intervals.

The above mentioned are just theoretical possiblilty. Whether it is practically viable for your situation, only you can tell. It may not work if the db is updated all the time.

From your post, I gather that your current provider do not have good uptime. In this situation, I would move to a better provider that has very good network/uptime. This would remove the concern of data integrity, managing and updating 2 servers.......

[Blapto]

No provider can guarantee 100% uptime, it's just not possible.
Your website is only 30MB. Let's say 50MB with future expansion. Put that in two more or less geographically similar locations Florida and New York, say) and you could transfer a 50MB dump between the two in no more than 10 seconds. Use rsync and that would probably be cut down to 1MB changes/day. You could rsync every 10 seconds and notice no performance loss.

[TDMWeb]

If you want to have automated failover between two different datacentres (which is a very sensible move IMHO) then a software-based failover system would be appropriate.

www.DNSMadeeasy.com looks like a good bet, certainly cost-effective. Alternatively look for a host that will provide a service based on SimpleFailover.

As has been said before, round robin DNS will not give you failover, but load balancing.

Your key issue will be making sure that the databases on both machines are up to date.

[FHDave]

I have seen many people asking about how to implement fail over/redundancy fail over. Most of the times, they are asking this because they know they are not using mission critical network/hardware and thus they start to ask about how to implement half-assed fail over/redundancy.

Implementing complex fail over / redundany is not easy and may cost you quite a bit. It will be much easier if you host your server in a reliable network if networ uptime is your concern. Or to use true server grade server if your concernis hardware. The rest you just work with your acceptable uptime guarantee.

For example, if you worry about network uptime, then go with a provider with a proven record. Internap, for example, has one of the best uptime guarantee which they deliver. We have been their customer for 3+ years and we have onyl experience few seconds of network outage due to their problem.

If you worry about hardware uptime, then purchase/use a true server grade hardware. For example, server with dual power supply, with RAID5 setup with Hot Spare drives, etc. If you have a server with redundant power, you also want to go with a data center with at least N+2 redundant power. Again, we are hosted with Internap and over the last 3+ years, we are yet to experience a power failure affecting both of their redundant power.

This is as simple deployment as you can get, yet as high uptime as you can benefit. Of course, neither Internap nor true server grade server is cheap. But if you are hosting a mission critical project, then nothing is too cheap. We do not implement any fail over/redundacy mechanism, but our overal uptime has been better than 99.9x%. Does the addition 0.0y% uptime worth the additional complexity and additional cost by implementing fail over/redundancy? If so, then by all mean, do it.

Good luck!

[rasputinj]

for a low cost type of solution DNSMadeEasy does work well for Failover, I use it on a few of my sites. Downtime tends to be around 5minutes, and it seems most DNS entries get updated because of the low TTL time. I use rsync and and database dumps to keep my stuff in order. I have a lot less stress now with backup sites. When something goes down, I am rushing to get it fixed, or get to a computer and an internet connection to get things up and running.

[efarmer]

Well said Dave,

unfortunately some mission critical setup seems to be cost critical setup thus having worrying uptime.