[neps]

we currently have a G3 Server (Apple issue with the 5 ethernet ports) which is now running OS X Server. We are thinking about purchasing a Dual G4 processor to run OS X.

Currently our G3 is used for nightly backups and morning pickups. But thats about it. We want to be able to run files straight from the server so people wont have to back up at night, the latest files will always be on the server. We are worried about this because what happens when the server goes down (which happens a bit) and we are working off a file on the server? Does the file go caput? Can we recover without loosing all we have worked on the client machines? Do the clients keep going on their merry way with the server down and reconnect when it is back up?

Also, we want to set up a type of raid system, where when one server goes down, the other will kick over? So say for instance we have the one g4 raiding to the g3, then the g4 goes down, can the g3 automatically be the server that is the main server?

One last thing.. does the auto restart feature on the g4 work okay? We are having trouble with the g3, where the appletalk services wont work and need to kill them in order to get it up again, and if that doesn't work we get a spinning ball making us have to restart. (Of course this could just be because of the old hardware, or even because it is 10.0.4 and not 10.1... maybe it is something fixed int here) This auto restart feature doesn't restart a frozen process, just a frozen machine?

thanks

[bluedog]

The trouble you are having with your server is horrific. There is no excuse why a machine that is a server should be causing so much trouble.

At my work, we are running an older beige G3 300mHz with AppleshareIP6 and it rarely crashes (once in about a year) and runs constantly with few reboots (for maintenance) and we run and open and work on files from this server.

Its approximately 15 people running files (5 creative staff) that open/save/run ALMOST ALL of our files from the fileserver. Only some of the administrative and salespeople keep files locally.

You should do a clean install of the system and fileserving software on this machine. As far as wanting to automatically switch-over to another server. It won't happen "on-the-fly" as it CAN when you are dealing with web servers that don't require a persistent connection to a server.

RAID is a way to put several disk drives together to logically function as a single filesystem. It can be set up in different configurations according to what the desired performance is. It is NOT a way to connect several computers together. Some RAID configurations allow for redundancy and failsafe features where it will continue to function even when a single drive in the RAID array goes bad.

I would talk with a local sales rep for technical assistance or sales assistance, but your current machine should not be retired or relegated to 'second' position by any means.

[King Chung Huang]

At my workplace, the server commons.ucalgary.ca is an old beige Power Mac G3/333 MT running Mac OS X Server 1.2.3. It does everything, from acting as a QuickTime Streaming Server, a web server, serving databases, used as a dumping ground for mass file transfers, you name it. There's definitely something fishy about your G3 Server. A clean install of Mac OS X Server 10.1 would probably clear that up.

[geekstud]

Originally posted by neps:
<STRONG>we currently have a G3 Server (Apple issue with the 5 ethernet ports) which is now running OS X Server. We are thinking about purchasing a Dual G4 processor to run OS X.
</STRONG>

First you should consider your needs/requirements, and then you can match those to the appropriate hardware and software solution. A Dual G4 is a cool machine with plenty of CPU horsepower, but your fileserver requirements may point to less need for CPU and more need for a fast disk subsystem. Of course, if you have the funds, you can have the best of all worlds - a Dual 800 G4 with plenty of RAM, a multiple channel SCSI card(s) and fast SCSI disks.

<STRONG>
Currently our G3 is used for nightly backups and morning pickups. But thats about it. We want to be able to run files straight from the server so people wont have to back up at night, the latest files will always be on the server. We are worried about this because what happens when the server goes down (which happens a bit) and we are working off a file on the server? Does the file go caput? Can we recover without loosing all we have worked on the client machines? Do the clients keep going on their merry way with the server down and reconnect when it is back up?
</STRONG>

Consolidating your workgroup's files on a central server, as your thoughtful questions indicate, has benefits as well as some risks. From an organizational standpoint the central server makes good sense, but moving to that model brings with it the requirement to ensure the "all the eggs in that one basket" with be safe.

When a server crashes while clients are accessing a file several different things can happen depending upon the exact state of the file and what I/O operations would be occurring. In most cases the file will be unaffected on the server once it reboots/recovers. In most cases the client may experience a short "freeze" when the server becomes unavailable, but then should continue working. The Appleshare clients will have to remount the server filesystems when they are available. If the client stays running it could write any files that have been opened to the local disk for the short term.

Other centralized services, including user authentication, will of course be "down" as long as the server is down.

<STRONG>
Also, we want to set up a type of raid system, where when one server goes down, the other will kick over? So say for instance we have the one g4 raiding to the g3, then the g4 goes down, can the g3 automatically be the server that is the main server?
</STRONG>

Such a setup is generally called a "high availability" or "redundant server" configuration. RAID, which is an acronym for Redundant Array of Inexpensive Disks (or Devices these days), is a system with several models for utilizing multiple disks/devices logically concatenated into a single "disk/device" to increase performance and/or availability in the case of the failure of a member disk/device. A proper RAID setup is an important part of a well configured server, but except in "higher end" server setups (e.g. Sun/Solaris with Veritas) it isn't a part of linking multiple servers together in some form of failure setup.

I think Apple would do well to bring such advanced functionality as server failover into the workgroup server arena. There are, of source, "low tech" approaches to providing some level of higher availability servers. You could purchase two servers and utilize one as a "warm spare". By using external storage enclosures you could facilitate the possibly of a "hand swapped" redundant server (power both servers off, swap SCSI/Firewire disks to standby server and power it up).

Only you can determine the cost/benefit ratio of the investment. Regrettably, many businesses fail to ask and act upon a simple question: "for how long can we tolerate the server being down?" I always tell my clients that my job isn't to design a solution based upon everything working, rather to assume the failure of various components (e.g. a disk goes bad, a power supply in a server fails, etc.) and design the solution to meet the answer to the downtime question. The less downtime the client can tolerate, the more complex and expensive the solution must be to address those needs.

<STRONG>
One last thing.. does the auto restart feature on the g4 work okay? We are having trouble with the g3, where the appletalk services wont work and need to kill them in order to get it up again, and if that doesn't work we get a spinning ball making us have to restart. (Of course this could just be because of the old hardware, or even because it is 10.0.4 and not 10.1... maybe it is something fixed int here) This auto restart feature doesn't restart a frozen process, just a frozen machine?
</STRONG>

It's hard to tell what the issue might be, but I agree with other responding posters that a good clean OS install wouldn't be a bad place to start. There are some known issues with Mac OS Server - if you are running the older V1.2 release your symptoms seem to match a rather nasty problem known as the "Desktop DB" issue. Check the Apple support forums for a description and some solutions.

Cheers,

-Nathan

[rkt]

Originally posted by neps:
<STRONG>We are worried about this because what happens when the server goes down (which happens a bit) and we are working off a file on the server? Does the file go caput? Can we recover without loosing all we have worked on the client machines? Do the clients keep going on their merry way with the server down and reconnect when it is back up?</STRONG>

depending on the type of files & your workflow, it might be worth investigating a cvs based system. cvs allows you to store all your files centrally. when a user wants to view or modify a file the server checks out a local copy to them. when the user is finished they check the file back in. the server keeps track of which users are doing what, and makes sure that all copies of the file are kept up to date.

as the user only modifies local copies of the file, if your server goes down during while editing they are not affected. this also means you could quite easily develop a redundant system where 1 or more servers provide a backup system for your main server.

[ 11-09-2001: Message edited by: rkt ]