Mirroring a server (For automatic failover protection)?

[zacdl]

So SBS cannot be clustered.

So- what can I do to minimize downtime? I have secondary (and third) DNS servers in all client records pointed out to the Internet, so Internet downtime isn't much of an issue.

I have RAID5 with a hot spare- so hard drive failure isn't much of an issue to me.

Now, if the MB or something fries- that throws me a huge problem.
Is there any way I can have a "mirror" server running on the network to automatically take over? I know you can only have one SBS box, so I was wondering if there wasn't another way...

And if for some reason a virus sneaks through the system (NOD32 updates every few hours- so I'm pretty much as protected as I can get- but the point remains...) and wipes out my programs, I don't have much backup for that (Tape backup right now, but it only has a certain limit before we outgrow what can fit on a single tape).

I was thinking I could use TrueImage to create an image of the entire system every 2 weeks or so (I imagine it would take awhile to create... not sure on compression settings so I don't know how many disks it would take).

But that leaves data up to 2 weeks out of date.

Just suggestions, please

 

[da sponge]

When you outgrow a single tape you get an autoloader

You can have another DC, it just hast to be win2k3 and not SBS.

 

[zacdl]

OK, but it would take over all operations of the SBS in the event of a failure?

 

[MorfiusX]

SBS isn't designed for high availability. Any of the products I have seen and used for mirroring or high availability are more enterprise grade products.

 

[da sponge]

OK, but it would take over all operations of the SBS in the event of a failure?

No (just domain controller roles). As morfius said, SBS isn't a HA solution. You have to weigh the cost of a fully redundant system against the cost of downtime while factoring in the likelihood of a failure occurring.

What is SBS running? Exchange? For failover you need enterprise edition (times 2) which is very expensive. File sharing you can use DFS to mirror the contents.

 

[zacdl]

It is primarily running exchange.

I guess I can just pray nothing more than a hard drive dies- as there is just nothing to take over in the event of a failure.

I am thinking of implementing a NAS for the server to backup to as well as RAID 5 (In my eyes- it can handle the bulk of software backup, and tape can be used more for system items like AD, Exchange, etc). Anyone recommend an OS for it? I've heard of FreeNAS. It plays well with Windows? Any problems with it?
Does it come with a firewall of sorts? For instance, I would want to block out every single communication other than from the primary server (and I guess the WebGUI to manage).
If not, what's a good one?

How about Ubuntu Server?

 

[pigster]

I have RAID5 with a hot spare- so hard drive failure isn't much of an issue to me.

Having a single hard drive die is only one of many possibilites - consider that electrical noise (surges/lightning) or power supply failure could destroy your array in an instant.

Now, if the MB or something fries- that throws me a huge problem.
Is there any way I can have a "mirror" server running on the network to automatically take over? I know you can only have one SBS box, so I was wondering if there wasn't another way...

You could use a piece of software like Double-Take for SBS and a standby server, but that's a pretty pricey solution. But it would limit your downtime to minutes.

You could restore from a backup to a standby server using the same hardware as your main server, that would limit your downtime to 3-4 hours.

(That's my solution, one of my member servers uses the exact same hardware as my SBS server, so I could restore a backup to it if needed)

Or just use current, easily available hardware and rebuild the server as needed and restore from a backup - but your downtime may be a day or two.

In any case, you should really evaluate your backup strategy, a business should do a full backup at least once a week with incremental backups every night. Consider using external hard drives, they're fast, reliable and fairly cheap. You can also rotate them and keep the offsite so a fire or whatever won't cripple you.

 

[moetop]

Depending on load, budget, and applications you may want to look at VMware. Save a image every night on the backup VMware server (Overwrite the old one) and still keep normal backups. Booting it up in case of emergency will take as long as booting a system. (2 to 5+ minutes) If you buy into their high end stuff you wouldn't even have to touch anything, but with a little intervention you can basically do it for free ( VMware server is free) With a little creative scripting you may also be able to obtain similar results with VMware server

 

[zacdl]

In any case, you should really evaluate your backup strategy, a business should do a full backup at least once a week with incremental backups every night. Consider using external hard drives, they're fast, reliable and fairly cheap. You can also rotate them and keep the offsite so a fire or whatever won't cripple you.

Full each night (Only takes about 40 mins or so to tape right now...). But if the system crashes- all the programs have to be reinstalled, etc.

Which is why I was wanting to know how long something like TrueImage takes.
I cannot remember how many GB of usage I am currently using, but I am thinking if I can image C:\ once a month (or more often if I install some critical apps) that would take care of getting the SYSTEM going.
I can then have my Data (D:\) restored from external hard drive backups, or a NAS.

 

[dahlhana]

If you are serious about uptime, and clustering is not an option, you should check out Neverfail (http://www.neverfailgroup.com/home.asp). These people have a seriously cool product that updates a secondary hidden server constantly and once the primary server fails, the secondary takes over without any interaction or interruption. This can be done locally or over wan..

 

[MIRTBelGeran]

I'm goign to take this in a differant direction. What OEM is your server hardware. if it is home grown I would ditch it for a main hardware provider that can offer 24 x 7 hardware support. If you pay for a service contract you can usually get your server up and running from a total melt down in 4-6 hours.

 

[swatbat]

When you outgrow a single tape you get an autoloader

You can have another DC, it just hast to be win2k3 and not SBS.

Or a bigger tape drive. Then you go for the autoloader.

I'm goign to take this in a differant direction. What OEM is your server hardware. if it is home grown I would ditch it for a main hardware provider that can offer 24 x 7 hardware support. If you pay for a service contract you can usually get your server up and running from a total melt down in 4-6 hours.

I can't remember if the OP listed the hardware in another thread about his issues but yea this is a good point. Something like a dell 2900 server(enterprise grade) would be a good choice for a server with great support. We use them for a lot of are clients as they are only a few hundred bucks more then the lower level ones. Dual power supplys, teamable nics, hotswap trays for the raid controller etc. Hell if you really wanted to you could make the memory redundant but I don't know anyone that has bothered to do that.

Anyway OP if you are really worried about downtime get a full backup everynight and have a server configured the same sitting in a box. Server has a hardware failure you move the drives over to the other one and turn it on. Software gets fried you restore to the second box and get them back up then fix the main box and copy the newer stuff over.

To do clustering would be very expensive. The extra box would help limit downtime without the extra expense of the external raid box for data drives and enterprise versions of windows software.

 

[zacdl]

It's an HP Proliant- based on past hardware issues- they are rather lazy in getting anything shipped...

Neverfail looks nice. Is it a bit for bit mirror of drives, or just of certain folders?

 

[swatbat]

It's an HP Proliant- based on past hardware issues- they are rather lazy in getting anything shipped...

Neverfail looks nice. Is it a bit for bit mirror of drives, or just of certain folders?

What level of support do you have with them? I've had real good luck with hp in the past.

With dell(which my current company uses) we usualy try to get are clients to get 4yr responce. Last few times we used it we got the part in 2 hours. They were simple failures though. I'd expect a day to get a lot of the parts

 

[zacdl]

Neverfail looks nice, but it does not do SBS (if I mentioned that already??)???

My situation:

#1. Main server (DNS, Apps, Exchange, etc)
#2. (Future) Application server to run a single application (its sole purpose). Proprietry system.

Those are the two main servers, add to that my wanted backup servers:
#3. Data backup server. A P4 with large amounts of HD space- should be great for backing up the other servers and not much else. Pretty much protects against data loss.
#4. (Future) Standby server. To take over if either #1 or #2 fail.

Now, I just don't know how to implement that.
#2 has constantly changing data. Which almost entirely forces me into some sort of mirror situation. But I don't want the standby server to ONLY mirror #2. I want it to take over #1 if it fails as well.

My thoughts were using TrueImage. Works great for #1. Restore image- restore data off #3 to latest.
But with constantly changing data on #2, TrueImage doesn't quite work for me if #2 fails.

Any ideas? My main thing is I don't want to have two costly standby servers (both would have to be able to take a decent load) sitting around when I can just have one.