Take the 2-minute tour ×
Server Fault is a question and answer site for professional system and network administrators. It's 100% free, no registration required.

I have two servers Main and Backup, such that I want a Backup server to be able to replace the Main one as fast as possible without physically accessing them.

So I want the Main and the Backup server to see same hard drives. (Even if backup does not mount them). Also I don't want stuff like NFS and I don't have very complex stuff like SAN.

So ideally:

Main    Backup
  |\     /|
  | \   / |
  |  \ /  |
  |  / \  |
  | /   \ |
  |/     \|  
HD A     HD B

So Main is connected to HD A and HD B organized in RAID 1 and if Main crashes Backup mounts the same set HD A and HD B.

Also I expect that the connections would be fast and efficient (FB or stuff like that) and remote such that different disk and different servers may be placed in different enclosures.

I want this to be as simple as possible no central storage device connected via SAN.

  • Are there hardware technologies to do this?
  • If there are what are they where should like look from?
share|improve this question
    
As MDMarra said by alluding to "shared DAS", you can do this with plain ol' SCSI drives, as long as your drivers and OS support the shared access without stomping on each other. That pretty much limits you to pairs of servers though, and they obviously can't be too distant; you're limited to the max length defined by the version of SCSI you're using. And, this isn't very easy. You'd do much better to look into SAN of some sort - iSCSI is getting pretty affordable these days and will be much more flexible. –  mfinni Aug 2 '12 at 14:47
1  
This is really not a simple problem. It is full of pitfalls and adds its own failure modes. That's why working solutions (like SAN) are complex. One solution I have used in a few cases is running rsync to keep separate drives on separate machine replicated somewhat close in time (how often I choose to run rsync). This would be for data, not the OS. –  Skaperen Aug 2 '12 at 19:06
1  
What you were expecting from us is unreasonable. You've already been told that you're looking for an external SAS enclosure and SAS HBAs for the servers. @ewwhite even linked to a specific product. As for how to set it all up, that depends on your environment. You need to do some more research on this before you can expect us to give you and more help. You're still in the exploratory stages of this project. We've pointed you in the right direction. Now you need to go do some reading and come back when you have specific questions, not just "so, how do I do it." –  MDMarra Aug 2 '12 at 19:48

3 Answers 3

up vote 5 down vote accepted

You would need a device with shared direct-attached storage capabilities. In the past, with parallel SCSI, we used to deploy HP's packaged cluster solution: Two nodes, a heartbeat provision and shared access (via SCSI) to the same drive shelf. It's exactly what you were looking for.

A modern solution for this would be something like two nodes, but connected to a P2000 storage enclosure with SAS HBA's and cabling. On top of that, you'd need some level of a cluster-aware filesystem and some ability to communicate heartbeat/status between those nodes.

However, the focus is misguided. Technology has changed in that we don't focus on individual server health anymore. Virtualization, and the abstraction layer that comes with it, have allowed applications and services to scale horizontally. The chances that one system will fail in such a manner that the sevrer is rendered unusable, without affecting the other server, storage, surrounding environment, is slim.

Can you provide more detail as to what's running on the system and what you're trying to protect against?

share|improve this answer
    
Thanks. See I was looking something without storage overhead. Ideally just a disk that can be hotplugged to different computed instead of physically pulling it out from one and putting to other. I'm aware of storages but this is exactly I want not to do because it has its own overhead and I need the disk to perform exactly as it is a physical disk attached to the server. –  Artyom Aug 2 '12 at 20:11
2  
@Artyom First, the P2000sa looks and acts just like a directly attached disk. If you had read the product documentation you would know that. Why exactly is it necessary that the storage subsystem look like a directly attached disk? Are there additional requirements you haven't informed us of? –  Chris S Aug 2 '12 at 20:42
    
To prevent any overhead between the server and the physical disk. –  Artyom Aug 3 '12 at 3:33

So I want the Main and the Backup server to see same hard drives
...
I want this to be as simple as possible no central storage device or SAN

No. You'll need centralized storage, whether it's a SAN, NAS, or just shared DAS.

So Main is connected to HD A and HD B organized in RAID 1 and if Main crashes Backup mounts the same set HD A and HD B.

What you're referring to is generally referred to as active/passive failover clustering. You can do application-level clustering with things like Microsoft Failover Cluster Services or you can do OS-level clustering with things like VMWare vSphere Fault Tolerance (though HA is usually sufficient for most use-cases).

Honestly, it sounds like you really don't have a clue what you're asking about. Not to sound mean, but you should really consider hiring a consultant with experience in designing highly available infrastructures. You're in way over your head.

share|improve this answer
    
I'm talking about shared DAS. However how do I physically connect same HD to two different servers... –  Artyom Aug 2 '12 at 14:52
    
I mostly talking about not Centralizing stuff like buying a central storage systems but rather how do I make DAS that is shared –  Artyom Aug 2 '12 at 14:53
1  
You purchase a DAS that supports being directly connected to multiple external SAS HBAs. Then, you mount the volumes that you want shared and follow your OS or app's guide for clustering. Not every DAS supports simultaneous access to LUNs though. Why are you so against using SAN or NAS storage? That's the real, scalable, answer to your question. –  MDMarra Aug 2 '12 at 14:56
3  
Call a vendor and explain your requirements to them. Thats what sales engineers are for! You explain your problem, tell them what you're looking for, then they'll tell you what products they have that meet your needs and will usually even suggest configuration options based on your specific OS/applications. You're essentially asking for us to do your job for you...for free...which is insulting. We're here to help with specific questions, not be your Systems Architect. –  MDMarra Aug 2 '12 at 19:50

Like MDMarra, I'm concerned that you are rather out of your depth here.

You didn't mention what operating system this runs on, what protocols are being used to access the files, whether these are on the same site, how much storage you need, how fast the storage has to be, nor how fast the failover has to be, nor whether you want the failover to be visible to the users (i.e. whether you need address takeover) nor how much time effort and money you're willing to spend on this.

I don't want stuff like NFS

But by definition you are using a network file system if you have a file server. Or are you specifically refering to the NFS protocol?

A list of possible solutions would be enormous. But starting with the cheaper ones on Linux (since MDMarra has already mentioned Microsoft) you might want to have a look at DRBD + local disk in a mirror, keepalived, iSCSI, shared SCSI devices, GFS2 and AFS.

share|improve this answer

Your Answer

 
discard

By posting your answer, you agree to the privacy policy and terms of service.

Not the answer you're looking for? Browse other questions tagged or ask your own question.