Quick question: easiest imaging software for instanty backup

[Tea]

Simple question: what's the easiest way to achieve this goal:

Server will run W2K SP4, 80GB

At least one of the workstations will be identical on a hardware level.

What I want to do is fit both machines with a hard drive caddy, and have a daily or weekly backup run on the server to image the server drive to a spare 80GB drive.(This is in addition to the routine data-only backup to CDR that they already do.)

If the server falls over (the PSU fails, let's say), then, as a short-term emergency measure, they can pull the drive outr of the workstation and plug in the backup drive. Reboot, and hey presto! we have a clone of the server back on-line - and they have done it all without having to call me!

Later on, I can drop by and mend the server.

The question is, what software package should I use for this? Ghost? Drive Image? Something else? Anything else I need to know?

Thanks guys.

Small hairy one.

 

[flagreen]

I use Drive Image. I've backed up XP Pro, Win 2003, Win 2K and all the 9X versions with no problem.

 

[Howell]

Re: Quick question: easiest imaging software for instanty ba

Will the server run 2k server or wkst as its OS?

 

[Tea]

Just ordinary W2K, Howell. In fact, I don;t actually know what you canachieve with W2K Server Edition (aside from spend a vast amount of extra money, of course). Anyway, we are looking at a network of ... er ...

1: ADSL modem
1: Firewall/DHCP server (a Smoothwall)
1: Main server
1: Accounts machine
1: Reception machine
1: MD's notebook
2: CAD machines
1: assorted other duties machine
1: stupid HP all-in-one laser device on a JetDirect box which is a complete pain in the arse, but at least I know what hoops you have to jump through to make the POS function now.

Oh, there is a plotter somewhere on thge network too. I think it's attached to one of the CAD machines as a local printer. Seems to be trouble-free.

 

[Tea]

Oh, and Mr Flagreen: is Drive Image trouble-free? Or are there hidden issues? No problems with locked files or anything like that? I want to be able to set it up to simply image the whole drive over to the spare drive with a single click (well, OK, three or four clicks will be OK) and, if need be, simply plug the backup into the accounts machine (which will have identical hardware), boot off it, and have the server back up and running with the same drive assignments & etc as before. This is the sort of thing that their PR department says is the purpose of the software, but I want to see it confirmed from a hands-on user that the process actually works before I comit my customer to it.

Thanks for your help.

Tired and grumpy hairy one.

 

[flagreen]

I don't know the answer to your question Tea. I don't use DI the way you are intending to. I run it off of two floppies and it will copy or backup anything that way. I've not had any issues with it.

 

[Mercutio]

You want Acronis TrueImage. Everything else, you have to boot to not-Windows to actually make the image. TrueImage works while you're using Windows and can do a complete backup, including locked files, apparently (I've never tried using it that way).

 

[mubs]

The new version of DriveImage (2003?) apparently can image the Windows partition from within Windows while Windows is running. There's an easily found thread on SR about it.

But genereal impression is that people don't like it compared to DI 2002, and that it is very slow.

 

[Tea]

TrueImage looks good, Merc. Reasonably priced too. Thankyou.

 

[Tea]

Hmmm.... On second thoughts, it seems to be targeted at creating image files. We don't want image files that have to be restored, we simply want to copy the chole thing to another drive. Am I wishing for the moon?

 

[Tea]

Hmmm.... On second thoughts, it seems to be targeted at creating image files. We don't want image files that have to be restored, we simply want to copy the whole thing to another drive, ready to plug in and go. Am I wishing for the moon?

 

[mubs]

Looks like you want one of the "Drive Copy" utilities that clone a drive. Drive Image will do this. A crippled version of Drive Image that Powerquest calls Drive Copy can be found here. Drive copy is limited to 80GB drices, and runs in DOS mode, I believe :-?

I've seen similar programs at stores like Worst-Buy, Fries, etc.

 

[Mercutio]

There are two stumbling blocks here: One is that Ms. Hairy Aussie wants essentially RAID1 over a network.

To the best of my ability to determine, that ain't gonna happen. Most backup or imaging software wants to back up to an image file format of some sort (stumble), and most network backup software does a REALLY lousy job with the euphemistic "System State" info (stumble stumble).

The only thing that I can think of that comes close given the criteria in place is, don't laugh,

Iomega Quiksync + Automatic Backup.

The only thing I'm not completely sure about in that case is system state data.

 

[Handruin]

What about Veritas volume manager? Last I used this product; you could synchronize and split volumes through the use of their volume management utility, which is an expanded version of the windows volume management tool that ships with windows.

VxVM was a very nice tool and it even allowed you to migrate data in a RAID 5 setup if one of the volumes was overburdened...but I digress.

If I'm not mistaken, you can create a RAID 1 volume, synchronize the pair, and then split it, ultimately giving you two identical volumes. Now the question remains, can both volumes boot? That I can't answer, as it is unexplored territory for me. Theoretically I'd say yes because a RAID 1 volume should be able to survive if one drive dies.

Veritas offers a demo of their flash snap utility. If it is like prior version, it's might be a 60 day demo.

 

[Mercutio]

Both drives would be bootable, but it was my understanding that Tea's backup drive wouldn't be living on the same machine that was being backed up....

 

[Handruin]

Might I add that Veritas can synchronize and split the drives without rebooting and while production is occurring? I don't know what would happen if you disconnected the drive while the system was live, but I suspect it would work.

One problem I see you may face is that VxVM doesn’t support W2K workstation…only server/advanced server/datacenter.

 

[Handruin]

The way I read Tea's description is that there would be this "floating" drive. The drive would be placed into this one machine at a scheduled interval and then a synchronization would be done to make a point in time copy of everything. The drive would then be moved to the other machine if one died...

Quick summary as I read it:

Both machines have hard drive caddy.

Daily backup run on the server to image the system to a spare 80GB drive.

If FAILURE Occures:
1.) Remove dead drive
2.) Plug in backup drive to working server and presto!

That's how I understood Tea.

 

[Handruin]

Tea, if you're building this for a customer, you might want to suggest the N+2 scenario if the data or machine is that critical.

Have them buy 2x 80GB drives and rotate the backups. (backup to drive "A" on one day, drive "B" on the next) Or...install two hard drive caddies in each machine and you can have a 3 volume mirror using Veritas. Split them at different intervals and this will give you a rotating backup...

We use this concept at work with what we call BCV's. These are floating mirrors that perform full and/or incremental establishments. (Synchronizations) A customer with a sensitive database may have 5 to 10 of these per production volume and perform periodic incremental establishments and splits to create point in time copies almost every hour. Every hour copies to a new BCV (mirror).

 

[Buck]

If the PSU or motherboard fails, you don't need a duplicate copy of the drive, you're just going to move the drive from one system to the other. If the drive or ATA controller fails, all you need is a mirrored set (RAID 1) in the same system for your duplicate information. In order to solve both of these problems, just have two identical server machines, with one hard drive each, and create a cluster. Is this possible, oh IT wonders of the world? I've seen it done with webhosting systems using UNIX, so it should also be possible with Windows Server. It may be more expensive, but I think Tea's present approach is far too complicated and difficult to manage. Plus, it is always best for this type of problem to be resolved without any end-user intervention (taking a drive from one system to another is the safest route either).

 

[Handruin]

Without stealing your idea buck, I was going to mention using MSCS, but the cost of a unified storage solution may be far greater cost than the simplicity of Tea's backup method, so I didn't mention this.

In a sense this person could benifit from a windows server 2000 cluster running as a two-node active/active setup where both nodes run production. If one node fails, the other node will perform double duty, but they will remain online.

Now the tricky part is that the unified storage has to be backup up either physically like Tea suggested or with the aid of a tape drive. The next level is to buy an enterprise class array which I assume is out of the realm of this project. Then the array can manage the backups, but the cost would be far greater.

I've worked with MSCS for the past 2 years and I've only dealt with array arbitration using fibre channel and switched fibre channel. I believe Tea would have to move to a SCSI system with an HBA capable of arbitration. That in itself would raise the price far above a simple 80GB EIDE drive. Maybe 3ware has a solution for clustering, I haven't checked.

 

[Buck]

Then how about having two identical systems except for the drives. One system has no drives, and the other system has two, in carriers that create a mirrored set. If the drives are okay, but the system is down, pull out the drives, and put them into the other system. If one of the drives fail, the system will still work with the drive left over. Replace the drive and rebuild the array. Better yet, have a spare handy.

 

[Handruin]

The missing answer is if both of Tea's machines are intended to do production work. If not, I think your idea is almost perfect.

The slight advantage of having the "floating" drive (daily backup) is that it will be removed from the system in an event of a virus or some other data corruption where a RAID 1 fails to protect from.

 

[Buck]

To me, the level of paranoia being expressed about the many different ways the system could fail (PSU, drives) and the data could be lost, warrants the extra cost. However, I personally would try to reduce those fears, and instead build a single system with RAID 1, redundant power supplies for the extra mile, and good anti-virus software like NOD32. The regular data backup to CDR provides the inexpensive offsite data storage if necessary. In addition, Tea’s boss should run the system through some burn-in time to verify items that fail less often, and if so, at the beginning of their life, such as RAM and CPUs.

 

[Handruin]

I guess we'll have to wait and hear about Tea's risk assessment and budget. What role will this server play? (database, file server...etc?)

If it has any kind of database, RAID 1 isn't the complete answer because of the fear of corruption. From the looks of things, it seems more like a file sharing server because microsoft will not let you install SQL server or Exchange on win2K pro. (If I remember correctly...can anyone confirm?)

 

[Tea]

Hoolie Doolie, now we are getting into deep water!

Doug expressed my idea well, I think. This isn't a mission-critical system, but they want to do it right - as reliable as possible within reasn and without going overboard.

Essentially, we want to be able to survive the failure of any individual machine and still do business while they are waiting for me to come along and fix it. If any of the other machines fall over, it's no big deal. One of the CAD machines, for example - no problem, there are two CAD machines, just use the other machine for a while. If both draftsmen are working, they will have to take turns at it. They can live with that.

Any of the other machines, the same applies. The key vulnerabilities are:

1: printer. You can live without a printer for a day or so.

2: Smoothie: you can live without the web for a day or so

3: Server. You can't do anything without the server. Ergo, we need a spare server that the customer can switch to if need be. Within, say, an hour or less. They are pretty bright people, but we want to keep it simple just the same.

OK, here is another idea. How about I build the server as usual, but split OS and data onto seperate discs? Then I clone the boot drive, and ... er ... I better think about this a little harder. Hang on, I'll ask Tannin. Back shortly.

 

[Handruin]

Depending on what the server does, would NAS be a possible solution?

The redundant server solution is one and a half steps away from a cluster. The cluster would allow production to continue in the matter of minutes, once the failed node no longer responds, the surviving node will take ownership of all resources automatically (if configured this way).

In the same breath, you would have a virtual IP address so no network reconfiguration would be required. The IP address is transferred to the surviving host. The single largest problem to face with a cluster is unified storage. (I'd have to research available options, but I suspect they are costly)

During the short failover sequence, the data may be unavailable for roughly 10-15 seconds. So the end question remains, is the cost of a cluster worth the increased availability? Could this customer use a Linux cluster? (I have no experience in this area) If the acceptable downtime is <2 hours, then your current solution seems reasonable.

I highly suggest that a failure scenario be run to verify the solution works.

 

[Tea]

Damn it, Tannin has been talking to the customer. I hate it when he does that.Throws all my grand pland out of whack.

David (the MD of the firm in question) has a notebook. He uses a simple routine: every morning, he synchronises his user files with those on the server, using Windows Briefcase. Through the day, he works away and re-synchronises before he goes home.This way, if either the notebook or the server fall over, he can't be more than a few hours worth of work down the tubes - which is an acceptable risk.

So he asked if the whole firm could work like that. Users work with files held on the server as routine and, at the end of each day, use Briefcase to synchronise their folder on the server with their own local hard drive. (The server is backed up in the usual way to CDR from time to time. Critical files daily, most things once a week is OK.)

If a workstation falls over, the worker concerned can simply borrow a spare workstation and continue working while they get me (or whoever) out to fix it.

If the server falls over, they have last night's backups on their local hard discs, and can continue working for a day or two while I fix the server.

The quality of staff they have, I am very confident that they will indeed do the syncronisation thing.

In fact, this sounds to me like a very workable, practical solution. Cheap, simple, workable. I like it.

Can anyone see a problem here?

Currently, all the machines run Windows 98 except for David's notebook, which runs XP (which he doesn't like as, among other things, it refuses to work properly with the Hewlettt-Crapard all-in-one ). The plan is to standardise all the systems (including the notebook) on Windows 2000.

 

[Howell]

I don't like Windows briefcase. It has been unstable for me in the past.

FileBack-PC will do what you want for a price. You can even set it to do the backups "on file change" to keep a backup throughout the day.

 

[Fushigi]

1. Server: Hardware mirroring with the mirror drive in a caddy.
2. Backup server: Runs normally off a boot drive in a caddy.
3. If Server fails, move drive caddy to b/u server, boot. B/U server becomes main server but lacks disk protection. Once Server is repaired, rebuild the mirror.

This covers every contingency except HD in caddy fails. And in that case you're still running off the primary drive so you're still up.

4. Buy a DVD+-RW drive and do daily incremental backups to it using backup exec or anything else that's convenient. Also do periodic full backups. Use RW media and establish a rotation. A minimal example would be:

(Set 1) Full backup
(Sets 2-7) Incrementals
(Set 8) Full Backup
(Sets 2-7) Incrementals
(Set 1) Full, etc.

Normal user routine is to simply swap in the DVDRW media as needed & off-site the previous backup disks.

 

[Howell]

Then how about having two identical systems except for the drives. One system has no drives, and the other system has two, in carriers that create a mirrored set. If the drives are okay, but the system is down, pull out the drives, and put them into the other system. If one of the drives fail, the system will still work with the drive left over. Replace the drive and rebuild the array. Better yet, have a spare handy.

I think I like this system the best. In addition you could have an OS installation on another caddy so that the backup server could be used as a workstation unless it is needed for server duties.

When the old server has been repaired it could then be used as the workstation with no need to interupt serving duties again.

 

[Fushigi]

(Set 8) Full Backup

That would be:

(Set 8 ) Full Backup

Emoticon madness...

 

[Bozo]

Distributed File system. Two servers that are set up this way keep the files current on each other. Pcs on the network connect to them with a common (1) network address. If one pc goes down, it's invisable to the users. Both servers are available for use as workstations also. Don't believe clustering is needed.

I am just starting to investgate DFS where I work, so I don't have a lot of info. The MS site has tons of info.

Bozo :mrgrn:

 

[Mercutio]

dfs is awful for creating extra replication traffic, and requires 2000 Server machines to handle that replication.

It works OK for a few hundred MB of data but even on a LAN you don't want to push into GBs of data.

 

[Bozo]

We're looking into two servers with a gigabyte eithernet card in each box dedicated to traffic between the two. If they are close enough to each other, a crossover cable will do.

Bozo :mrgrn:

 

[Mercutio]

That would certainly work.

One of my early consulting jobs involving Win2000, I tried doing dfs to keep a couple of servers up-to-date for around 6GB worth of files. It... did not work the way I had hoped it would.