Admin Alert: Things to Do When Adding Drives to a System
May 14, 2008 Joe Hertvik
My shop recently added 12 new disk drives to our i 550 development partition. While we outsourced the install to IBM, I found that there were a number of things I could do before and after the process to help the installation run smoothly. To share that knowledge with you, this week I’m presenting five things to think about when adding disk drives to a system.
Five Tips In A Nutshell
To help a disk drive installation run more like a gazelle and less like a hippo, I’ve found that the following five items can make a big difference.
Here’s how each of these items can affect a disk drive installation.
Convenient and Ample Downtimes
In some instances, you can install disk drives to a Power i, System i, or iSeries box without taking down the system. However, if you’re replacing existing drives, you will have to take down the system. This means that you will have to choose an installation window that is the most convenient for your users (or given the demands of 24-hour access, you may have to choose the least inconvenient time for your users).
If you need to take down your system, make sure that you budget plenty of downtime for installation. Unfortunately, you can’t usually count on installation being a two-, three-, or even a four-hour event. Some of the necessary steps may involve removing data from and reinstalling data on the system, creating a new RAID set, and rebalancing existing data on the system. For timing the install, a good rule of thumb is this:
It takes as long as it takes.
For my install, I was sure that it would be finished in four to five hours. Instead, it took eight hours because of some loooong system processing when creating a RAID set. Assume that installation will be a longer process than you think, and budget your downtime window accordingly.
Back Up Your System… Twice
Always perform a full system backup before you add or remove disk drives. If something goes wrong, you will want to restore from the most current copy of your database.
However, you’ll want to do more. You’ll want to be paranoid. Take two system backups. Please.
You want two backups to cover yourself in case there’s a media problem with the first set of media you’re using during a restore. If that happens, you can easily pop in the second media set and keep on restoring. If you only have one backup media set, you’re in trouble if there’s a problem with the media. So be paranoid and take two full backups before you start. You may not need the second set but it’s a good insurance policy in case the unexpected happens.
Do You Need To Back Data Off Your System?
A whole other layer of complexity enters the process if you are replacing drives on your system instead of installing additional drives. In our install, we replaced nine 35 Gigabyte drives with 12 141 Gigabyte drives. Here’s where our issue came in.
Before the install, system disk usage was at 89 percent but we were replacing and upgrading nine drives that contained 15 percent of our data. Furthermore, nine of the 12 141 Gig drives had to be installed in the same slots as the nine 35 Gig drives they were replacing. There were only three available slots for installing the new drives. Since we had to pull the old drives before installing the new drives, the data on the nine pulled drives wouldn’t have anywhere to go (i5/OS will move data from removed drives to open drives in the system, if there is space available).
This situation left us with two alternatives. We could remove 15 percent of the data from the system before we took off the 35 Gig drives (allowing enough extra space to move the data on the removed drives elsewhere on the system) and then re-add it back to the system after the 141 Gig drives were installed.
Or, our second alternative was to install three of the 141 Gig drives in the three slots that didn’t currently contain 35 Gig drives, and let the system move the 35 Gig data to the rest of the system, including the three 141 Gig drives, as the 35 Gig drives were removed. The risk with installing the three 141 Gig drives was that they had to go into the system unprotected without being part of a RAID set, and we would add them to the new RAID set containing all 12 drives after the other nine drives were installed. With the second alternative, the data would stay on the system and the data wouldn’t need to be unloaded and reloaded as we would have had to do with the first method. We went with the second alternative.
The morale here is that you may not always be able to add new drives to your system, without having to reload at least part of your data. Sometimes you will have to replace smaller drives with larger drives, and that will require you to think about how to perform the replacement and still keep all of your data. So before new drive installation, come up with your strategy for handling data migration.
Borrowing Processor and System Memory
If you’re adding disk to your system, one of the things that will affect your installation time is how much processor and memory are allocated to your partition. If you’re on a multi-partition system and you’re adding disk during off-hours, you may be able to speed up the installation by moving additional processor and memory to the system from another partition on your box. Perform an analysis as to whether or not the sister partition will miss the processor and memory you’re moving over.
If you do move processor and memory between systems, make sure to note how much processor and memory the system had before the move. This will come in handy when you’re restoring partition resources later on.
Clean Up After Yourself
Make sure that any temporary changes that you made to facilitate the install are reversed after the install is finished. This includes any borrowed processor or memory as well as restoring any data that was taken off to accommodate the install. The good news is that once you have the drives installed and you’re all cleaned up, there shouldn’t be anything else to do. Your system will be upgraded and your users will immediately be able to start using the new disk drives.
Client Access Mysteries, No Prizes, and Rick
In an April 30 article, I introduced my readers to Rick, who had a problem starting more than 70 Client Access sessions on his OS/400 V4R4 system. I wasn’t able to help Rick, so I sent a plea to the readership to come up with solutions. I offered the first reader to solve Rick’s issue an Admin Alert No Prize, modeled after the fabled Marvel Comics No Prizes of the 1960s.
Being a humble writer, I thought I’d get two, three, maybe five answers at best. I was wrong. I received at least 50 different entries into the Help Our Rick contest. I was blown away by the sincerity and logic that all of recipients put into their answers. As I promised, I also forwarded all the answers to Rick. As of this writing, I haven’t heard from Rick yet to see if anyone has solved his issue. As soon as I hear from him, we’ll announce a winner. Until then, thanks to everyone who entered the contest and tried to help Rick out. Hopefully, we’ll hear some results soon.
Correction to ObjectConnect Article
In my March 12 article about Using ObjectConnect to Move Objects Between Partitions, I wrote about how you needed a System Network Architecture Distribution Service (SNADS) network to use the product. Thankfully, I was wrong. Reader JRP wrote in to tell me that “..No SNADS connection is necessary for ObjectConnect to work.” He’s right. You can configure ObjectConnect to run over a number of different protocols and in IBM’s System Administrator Companion redbook, Big Blue stresses that ObjectConnect/400 is a set of six CL commands designed to move objects, libraries, or even integrated file system directories from one system to another without the need to implement a SNADS solution. My apologies for the mistake. If you want to find out more out about OptiConnect and how to set it up on your system, check out the iSeries Backup and Recovery, Version 5 manual (SC41-5304-06).