A lesson in upgrading hardware

The other day, I got a note from a longtime reader (I won’t reveal their name as it might be embarrassing for them) about a hardware upgrade to a NetWare server. In a large network, with many servers, this is a fairly straightforward task. But in a single-server network, such as a small business network, the process can be fraught with danger.

The actual process of upgrading hardware is relatively simple:

1) Make two complete and full, verified backups of everything on the server.

2) Down the server.

3) Replace the hardware and reboot.

If everything goes well, there are no worries. If a problem occurs, you can always bring in new hardware, install NetWare and then restore from one of your backups. But can you do a full restore?

We all know that you can’t back up the directory using standard archival tools. You can’t do it even with non-standard ones. In a large network, that isn’t much of a problem. Simply remove the directory replicas from the server you need to work on (making sure, of course, that other replicas of those partitions exist elsewhere in the network), replace the hardware with the three steps listed above, then push directory replicas back out to the new hardware. Easy-peasy, right? If there’s only one server, though, there’s no place to “park” the directory, is there?

Novell does offer some guidance (see https://www.novell.com/documentation/ndsedir86/taoenu/data/a5sq6ju.html for example). There is an option in the NWCONFIG utility called “Save Local DS Information Prior to Hardware Upgrade” (this name may vary depending on the version of NetWare and eDirectory you have). This dumps the information needed to reconstruct the datastore on that server as a series of files in the folder SYS:SYSTEM$HWNDS.BAK. You can then copy the files to a safe place and proceed with your upgrade. As always, be sure there are two, complete, verified backups of this data. By the time you discover that your only back-up copy is damaged, it’s usually too late to get another one.

Before restoring the data created by the NWCONFIG method, it is important – EXTREMELY IMPORTANT – to ensure that the new or upgraded server has the same tree name, address, and server name as the one you took the backup from. Especially when replacing a server – or replacing the disk drive holding the SYS: volume – it’s easy to overlook this requirement. If you do overlook it, though, you risk having to re-enter all of your data, manually. How many days will that take you?

If you take the time to read all of the steps needed, your upgrade should go smoothly (although maybe not quickly). Cutting corners to save time is not recommended. One reader has learned a good lesson, hopefully you can benefit from it.