Like many of you, I’m an NT professional. Maybe you can learn from a recent debacle I experienced.
I work for a company in the Toronto area, and my company’s main line of business is insurance brokerages. Lately, we have been dealing with network installs that take anywhere from three days to eight weeks to complete.
The conversion gone wrong
The worst install I’ve run across involved a broker’s management software (no names mentioned), which included all the names and addresses of every client, all of the company’s policy information, accounting data, and transactions with other insurance companies. Needless to say, this software is a business-critical application.
When we perform conversions and installations, we find ourselves moving most clients from a UNIX/VAX/XENIX platform on various types of servers where all existing hardware is set up for serial connections to dumb terminals and serial printers (fun, huh?) to a Windows NT/9x client-server network. All printers still run through serial ports, and the software is still UNIX-based.
This particular installation was brutal. It consisted of 50-plus seats. Once the PCs were installed, the server was implemented, and the network was made to function happily with all of the non-critical applications running on the new platform, we decided to continue with the UNIX-to-NT software format conversion.
The main problem to overcome was the need to get backup tapes from the old system restored onto the NT server. Doing so required help from the software publisher to get the backup tape database files converted and replaced on another tape for the new server. Of course, as soon as you start the backup process, no one can make any changes on the old system.
Users interpreted this to mean that their network was down. A backup on this particular system took about four hours to complete. After waiting for the company to convert its data offsite, I drove back and stuck the tape into the new server and the tape got eaten.
Now I had 50-plus users screaming questions about when the new system would be up and running. The only answer I had was, "Hopefully soon," knowing darn well I was probably going to have to work through the night to get the new system running.
Once I finally got another tape created, I decided to borrow another tape drive "just in case” and installed it into the new server. The data was soon restored, and the system was up and running.
Next was instructing the onsite administrators how to conduct backups and reboots, and all the other pertinent information. The process began with three people crowded into a closet where their server resided. We were all there when the NT server began experiencing problems.
After an hour of trying to determine the problem, I isolated the trouble as resulting from a software problem with the NT server and called Microsoft's tech support line.
I was given two options. The first involved a no-guarantees, 12-hour solution, while the second involved recovering from a system backup. The only thought I had was, “Why didn’t I conduct a system backup before going live?” Having been under the gun, I'd neglected that step.
Needless to say, the customer was down for three days and wasn’t happy, as we had said downtime would be 12 hours.
The moral of this story?
Never forget the fundamentals when working with complex installations and conversions. As tempting as it may be, don’t allow yourself to be lured down shortcut paths that lead to traps.
Also, don't give your customers a definitive time frame for service restoration, because things go wrong. Do conduct a complete system backup before unleashing the server to the users. Make sure your tape drives don’t eat tapes, and periodically check your backups to ensure they are working properly.
Did this NT professional make rookie mistakes, or have you had similar experiences as an industry veteran? Share your opinion by posting a comment below, or send the editor an e-mail.