|  |
Scripting Tools for SAN Volume Controller
An interface and scripting tools using Perl for automating tasks in SAN Volume Controller.
Date Posted: March 30, 2006
|
|
|
|
|
|
Please consult the SVC documentation for information on setting up SSH keys using ICAT. | | |
|
The demonstration script shows how a typical task that may be performed on an SVC cluster can be automated through use of scripting. This script demonstrates both usage of the library and how a Managed Disk Group in an SVC can be rebalanced to improve I/O performance. | | |
|
When a Managed Disk Group is originally created, it is generally created with a small starting set of Managed Disks, from which the initial Virtual Disks are pooled. Over time, the requirement for storage may grow, so additional Managed Disks will be added to the Managed Disk Group. Virtual Disks are then expanded or added.
Virtual Disks will initially attempt to evenly distribute themselves amoung the Managed Disks provided. However, when this storage is full, the new Virtual Disks will only fill empty space, such as the recently-
added Managed Disks. Therefore, when I/O operations are performed on the new/expanded Virtual Disks, I/O hits will be unevenly distributed between the physical storage media.
For the purposes on the demonstration script, a "balanced" Managed Disk Group is considered to be one in which all Virtual Disks are evenly distributed between the Managed Disks, weighted by capacity.
| | |
|
When run in the default, non-executing mode, the script will not run any commands against the server that will adversely affect SVC performance. The user can review the output of the script in this mode to execute migrations in a controlled fashion.
Running a large number of extent migrations in parallel has the potential of impacting the performance of the production workload because the SVC cluster will issue read and write commands to the storage controllers, consuming resources on those controllers. It is suggested that the number of concurrently executing migrations be kept small until the impact on the production workload has been assessed.
Before starting the migrations, ensure that the managed disks onto which the vdisks will be migrated are online and operating normally and that there are no unfixed errors in the error log relating to those MDisks.
In execution mode, the script will attempt to execute a number of migrations up to the point at which 32 concurrent migrations are in operation. This number can be restricted through use of the --number parameter. While the number of migrations is greater than or equal to this number, the script will not run any additional migrations. In this way, the potential impact of the script running in execution mode can be limited. | | |
|
The script has two modes of operation: The first simply puts out a sequence of migration commands that will improve the balance of the target Managed Disk Group. The second will automatically execute these commands. The script is also able to wait for these migrations to be completed before executing further
migrations. | | |
|
The script is design to automatically ignore sequential and imaged disks in all circumstances. If you require the script to balance sequential disks, thus converting them into striped disk, please manually convert them before running the script. | |
|
|
|
|
| |
|
