I have a production server on a Dell PowerEdge 2950 running Redhat 5 (Tikanga) and other apps for use with finincaial trading including LOMS (low latency order management system)

I need to make an exact copy of this server and all of the software running. Someone in my organization went and purchased a dell PowerEdge R710 as they were told the 2950s are no longer avail from Dell and are the closest thing hardware wise to the 2950. (I personally would have searched third party vendors to buy a New or Refurbed 2950 but I wasnt told about it until after we received the hardware...)

I know there will be some issues due to the fact that the hardware is not exactly the same. My questions are...

What is the best way to clone the original production server?

Is it possible to clone it without bringing it down?

Is it possible to do it via ethernet since it is in the same rack?

If I have to bring the server down it will have to be done on a weekend which I am trying to avoid

What can I do after it is cloned to get it working properly with the new hardware?


Thanks guys!

There are a number of options for this, I'm just going to run through a couple scenarios:
You didn't mention if this is a total replacement or just an addition, so I'm going to assume overlap.

1) Install + rsync: (best for overlap)
Install the new server using the same guidelines as the previous server (Whatever you guys tend to use), then rsync all the user shares/mounts/etc across. This way the system can be easily kept in pretty close step with the other system.

2) Full backup + restore: (for total replacement)
Use your last full baremetal type backup and attempt a restore. Patch it as needed to make it work on the new box, taking special care with things like /etc. This is not a preferred option, since I think you'll have more trouble with this than the other method.

In practice, what we do here is essentially option 1 for bare metal boxes. 90% of our systems are actually virtualized via VMware, so that makes cloning a snap. I assume you're talking bare metal, however.

The idea is if the main server dies for any reason we can just power on the replacement server and we are back up and running.

Oh - ok. So an HA/failover sort of thing.
Rsync may be the simplest free tool, but it won't be up-to-the-second accurate. Not sure what's in your budget, but you might want to seriously consider solutions like Veritas Clustering and RedHat Clustering:
http://www.symantec.com/business/cluster-server
http://www.redhat.com/cluster_suite/

There are also some others I haven't used:
http://www.linux-ha.org/HomePage
http://www.radiantdata.com/

Thanks for the info, if Rsync is not up to the second accurate, how often can it sync the two servers or how accurate would it be?

How much of a CPU load would there be using rsync? I say this because this box is responsible for low latency trades around the clock as we deal with international markets so we dont want to bog down the server while its sycning.

Hi,

Maybe I misunderstand this post but if you want High Availability and synchronize two servers then they both have to be up. In no way the second (fallback) will be up to date if you shut it down for whatever reason.

I use High Availability as pointed to in a previous post (http://www.linux-ha.org/HomePage) with MySQL database master-master replication and Unison for file synchronization. In my opinion Unison is a lot easier to setup and maintain then rsync (I used both). I'd have to look into the Unison manual to see if there's a way so synchronize on less then a minute time laps. I have it set to a minute. Unison uses very little resources and synchronizes flawlessly bidirectional (like in load balancing environment where you have changes made to both nodes of a cluster).

At any given moment when one of my nodes is down the other one takes over without a problem without our users even noticing any downtime.

Kind regards,

Eric

This is exactly what I am trying to do. I wouldn't need it to be accurate by seconds so minutes would be fine. How can I learn about achieving this kind of high availability where if our main server goes down the second node will take over?

Hello,

If you want the easy to follow step by step guide go to HowToForge and type High Availability in the search box. You'll get a lot of tutorials, most of them for Debian. It should be easy enough to 'adjust' the installation command to suite your OS.

If I recall correctly I didn't even have to reboot the servers to get HA up and running.

I have my systems running on Debian without any problem and used guides from that site. Whatever problem you encounter I'm sure you'll get an answer here at LinuxQuestions.

I would advice you though, if you're in a production environment to first try out the installation on a spare server or in a virtual environment if you have that available. If you want I can provide you the installation steps I followed to get our intranet with MySQL backend on High Availability with load balancing and MySQL master-master replication running. Just PM or mail me if you want that document.

Kind regards,

Eric

1 members found this post helpful.

Now the replacement server is not set up. I would need to initially clone the original production server that way the operating system and more importantly the LOMS application and MYSQL setup would be an exact replica as I dont know how to install this application from scratch or configure it. Would unison be able to copy it this way so I can have the application set up and installed exactly the way it is on the original server?

Hi,

Unison is a file synchronizer with a lot of options and possibilities but I don't guarantee that it's capable of 'cloning' a system that way since it wasn't constructed to do that. I think your best option here is to setup a CloneZilla server on another server or boot your server from a CloneZilla or Acronis CD and save a complete image to an external HD or something.

After that you can restore the image onto the new server just by booting it from the same cd CloneZilla or Acronis and restoring the image. After restore check if everything works or if you need to install some distinct drivers and off you go. Then you can set up the HA and replication.

Kind regards,

Eric

After researching clonezilla it looks like I can definatly use that to start the cloning process. Now once I clone it to the new box, how do I work aroun the issues of the different mother boards / chip set? The original server is a Dell 2950 the backup is a Dell r710

Hello,

I'm not that familiar with RedHat so probably not the best reference, but I believe that when you boot the 'new' server for the first time an automatic check gets performed every time. New hardware will be detected and if necessary you'd be prompted to take action and configure it. If I recall correct the hardware detection program launched at boot is kudzu. Most likely you can also launch it manually to check your complete system.

Kind regards,

Eric

There's a lot of tips in here, but seriously, if this is for production use you need to do some serious research on HA solutions.
Good HA doesn't 'just work', it requires planning, cabling, and well-thought out processes.

Ask yourself these questions:
1) What am I providing? DB connectivity? File shares? Proxy access? Email? Routing? You say it's for a LOMS/mysql, but you need to break it down to what you're actually doing with it. MySQL has clustering/replication built in, but it requires configuration.
2) What is the budget? Every solution costs money and or time. Decide how much of each you can afford.
3) What is the risk? 1+2 play into this, because the higher risk makes a huge difference in budget (or should)
4) How much latency can you tolerate? Having one machine powered off until the other dies will result in HUGE latency. Several minutes at minimum. True HA requires active/active type scenarios where there is dynamic routing and load balancing. Active/passive is failover and may be cheaper but often results in higher latency (60 seconds or more)
5) How much can you lose? Can you lose session persistence? Can you lose data? Can you lose files? Different scenarios result in different solutions (F5 load balancers, database clusters, failover networks, ad nauseum)

Everyone has posted really good options, but it sounds like your requirements are still pretty fuzzy and you may be new to HA. I would highly recommend you spend some time doing research for your own benefit. There a so many solutions out there that the field can be extremely confusing at first.

MBybee makes a lot of good points.
If you really want minimal/negligible downtime, use a cluster.
In either case, it's a non-trivial exercise and would need proper testing (always the fun bit!)