Admin Alert: Moving i5/OS Resources on the Fly
Published: July 30, 2008
by Joe Hertvik
With the newer Power i, System i, and i Series machines, IBM provides the ability to reconfigure your CPU and Main Memory (RAM) configurations on the fly. This week I'll show you how to set up your system so that you can automatically transfer CPU capability and memory between systems. It's not difficult, as long as you have the right version of the Hardware Management Console (HMC).
A Framework For Moving Resources
The best way to teach is by example, so let's create a framework scenario where I can illustrate the concepts and techniques involved. Let's say that our example shop has a System i 550 running i5/OS V5R3. The machine has two partitions (a production and a development partition), three CPUs, and 32 GB of main memory. The CPUs and the RAM are divided up between the partitions in the following way:
- The production partition has two processors and 20 GB of memory
- The development partition has one processor and 12 GB of memory
The development machine is busiest during the day, and it doesn't get much use at night. With foreign customers, a second shift, and an end of day batch procedure, the production system works around the clock. Production is so busy that the company wants to borrow spare CPU and memory from the development machine at night, so that the configuration profiles change to meet their production demand. To meet nightly demands for batch processing, the operations manager would like his partitions to look like this from 8 p.m. to 8 a.m.
- The production partition has 2.5 processors and 28 GB of RAM
- The development partition has .5 processors and 4 GB of RAM
This scenario requires us to perform two sets of transfers between the development and production partitions. One operation will transfer .5 processors and 8 GB of RAM from the development box to the production box at 8 p.m. A second operation will transfer the resources back from production to development at 8 a.m. the next morning.
I'll discuss how to automate these transfers on a nightly basis.
Part One: Checking Your Pre-Reqs
To run this automation, you need to be controlling your i5/OS system through a Linux HMC version 6 or above. I tested this setup by using HMC version 6.1.3, build 20080112.1. If you're using a pre-version 6 HMC, the options I'm describing here may not be available. To check your HMC version, start your server and open the Licensed Internal Code Maintenance→HMC Code Update node in the navigation area of the HMC screen. Your HMC version information will appear in the right-hand pane.
Before you can move processors and memory between partitions, you also have to make sure that your partitions are set up to send and receive the resources that you'll be shifting around. For processors, each partition contains a minimum, maximum, and desired number of processors that can be assigned to that partition. The same goes for memory. If the amount of resources that you're going to shift between partitions makes one of the partitions go over or under its maximum/minimum limits, the transfer will not occur.
To check what your maximum and minimum limits are, expand the Server and Partition→Server Management→Partitions→Partition name node to reveal your partition profiles. In the HMC, there are two profiles for each partition and both profiles usually have the same name. Highlight the partition profile name that is preceded by a check mark, and right-click on that name. Select Properties from the pop-up menu that appears. On the Partition Profiles Properties Screen, select the Processors tab to view the partition's processor properties and select the Memory tab to view the partition's memory properties for that partition. Here you can modify the maximum, minimum, and desired amount of processors or memory. Make sure that you set up boundary limits that match the desired configurations that you want to run on your partitions. Click OK to save the changes. The only trick is that you'll need to completely shut down the partition you're changing limits for and restart it for the new settings to take effect. Processor and memory limit changes will not be changed until the partition is reactivated from a cold state.
Part Two: Setting Up the Automated Transfers
Once your partitions are correctly configured, you can set up the transfers by using the HMC's Schedule Operations function. In the Navigation Area, open the HMC Management→HMC Configuration node. The right hand pane of the screen will show all the HMC Configuration options. Click on the Schedule Operations item, which brings up the Scheduled Operations window. This window will contain one item for each server and partition that is controlled by the HMC. If you want to transfer processor or memory from one partition to another, highlight the entry for the partition that you want to transfer the resources from and click OK. This will bring up the Customize Scheduled Operations window for that partition.
This window allows you to set up three types of scheduled operations for a partition:
- Activate on an LPAR--This allows you to set up a timed operation that activates an i5/OS partition at a scheduled date and time
- Dynamic Reconfiguration--Used to schedule a timed operation where a system resource (processors, megabytes of memory, or I/O cards) will be moved from one partition to another
- Operating System Shutdown (on a partition)--Allows you to schedule a date and time when the HMC will shut down a specific partition
To set up a resource move from the Customize Scheduled Operations window, click on Options→New from the menu bar, which will bring up the Add a Scheduled Operation window. Select Dynamic Reconfiguration from that window and click OK. This will bring up the Set up a Scheduled Operations window, where you can configure your scheduled move. This window has three tabs that allow you to configure the following pieces of a scheduled resource move.
- Date and Time allows you to enter the date and time when the transfer will occur. This tab also contains a Time Window area, where you can specify a maximum amount of time the HMC should take to perform the operation before it terminates the transfer.
- Repeat lets you specify whether the transfer is a one-time operation or whether it should be repeated on a regular basis.
- Options is where you tell the HMC which resources should be transferred (CPU, memory, or I/O adapter), which partition to transfer the resources to, and how much of the resource (if applicable) should be transferred.
For our example, where we want to transfer .5 processors and 8 GB of memory between our development and production servers each day from 8 p.m. to 8 a.m., we would enter four recurring transfers under Customize Scheduled Operations.
On the development partition, we would enter two operations: one to transfer the CPU and the other to transfer the memory. We would use the Date and Time tab to designate that each operation goes off at 8 p.m. We would then use the Repeat tab to specify that these are repeated scheduled operations that will occur on specific days of the week (Monday through Sunday) either for a specified interval (measured in weeks or number of repetitions) or infinitely (which means that the operations will continue running indefinitely until they are cancelled). On the Options tab, we would specify that we want to move the .5 CPU or the 8 GB from the development to the production partition. Once the two development partitions operations are entered, all the scheduled operations for the next week will be added to the Customize Scheduled Operations screen.
On the production partition, we would enter another two scheduled operations that send the resources back to the development partition. These operations should be scheduled to run at 8 a.m. every morning under the Date and Time tab and they would mirror the operations on the development box. Like the development operations, all the iterations of the scheduled operations for the next week will automatically appear on the production partition's Customize Scheduled Operations screen.
If we want to view the details for a scheduled operation, highlight the operation on the Customize Scheduled Operations screen and click on View→Scheduled Details. This will show you the details of how the operation is set up. The biggest pain about setting up automated operations is that you cannot change the operation from the Scheduled Details screen, you can only view it. If you make a mistake and want to change your operation parameters, you need to delete the operation (using the Options→Delete function on the screen) and recreate it again.
Rules for Transferring Memory Between Machines
When setting up a memory transfer between partitions, beware of these limitations:
- Memory is transferred from the *BASE storage pool of your source partition to the *BASE storage pool of the target partition. Before the transfer occurs, you must have enough memory in *BASE on the source machine to effect the transfer. If you don't have enough memory in *BASE, the transfer will fail. Be aware that it may be tricky to ensure that enough memory exists in *BASE, if you are running dynamic performance adjustment on your system, which automatically changes your storage pools to match system needs.
- Watch your partition's minimum and maximum memory requirements when performing a transfer. If the source partition's total memory will fall below the partition's minimum memory setting after the transfer, the operation will probably fail. Similarly, if the target partition's total memory will rise above the partition's maximum memory setting after the transfer, the operation will also probably fail.
- Check your minimum and maximum storage pool sizes for the *BASE storage pool on your source partition. If the transfer will cause the *BASE storage pool to fall under the pool's minimum storage on the source machine or rise above the *BASE pool's maximum storage on the target machine, the transfer will fail. To check what minimums and maximums are in effect for *BASE on both machines, run the Work with Shared Pools (WRKSHRPOOL) command from a 5250 session and press F11=Display Tuning Data to display the pool's limits.
If you check these simple settings, your transfer will be more likely to succeed. It's easy for a scheduled operation to fail if these values aren't correctly set.
Checking Your Transfers
To see if your scheduled operations completed or failed at their appointed time, open the HMC Configuration options again (by clicking HMC Management→HMC Configuration from the navigation area) and open the View Console Events option. This will provide a list of all the events that have recently occurred on the console. There will be entries for your scheduled operations that tell you when the operation started, when it ended, and whether it was successful.
And that is how you can automatically transfer system resources between systems. If you follow these simple rules, your transfers should process flawlessly.
Two OS/400 Performance Adjustment Settings to Check
Post this story to del.icio.us
Post this story to Digg
Post this story to Slashdot