IRIX 6.5 » Books » Administrative »
Origin FibreVault and Fibre Channel RAID Administrator's Guide
(document number: 007-3715-002 / published: 1998-12-23)
table of contents | additional info | download
find in page
Chapter 2. Using the RAID GUI to Configure ArraysThis chapter explains how to use the RAID GUI to configure an array. It contains the following sections:
Using the Host Administration Window to Add ServersWhen you start a RAID GUI session and no server hostnames are in the ssmhosts file, the list of host identifiers in the window is empty. The Host Administration window appears, as shown in Figure 2-1.
(You can open the Host Administration window during an ssmgui session from the Storage System Manager window by choosing Select Hosts in the Storage Management window's File menu. In this situation, the list contains the hostnames of the servers with currently managed arrays. You can add or remove server hostnames from this list.)
To add a server to the list, follow these steps:
In the Host Identifier field, type the server hostname.
Click Add.
Continue to add server hostnames as desired.
Servers on this list are called managed servers.
 | Note: If you include the same server under two names (for example, as machinename and as machinename.domain.com), the GUI opens each as a separate host, which thus takes twice as long as necessary.
|
When the list contains all the server hostnames you want, click OK. A window opens asking if you want to save the new host list in the ssmhosts file in the home directory.
Register your choice:
Click Yes if the list contains only servers with storage systems that you want to manage in most future ssmgui sessions.
Click No if you added servers that you want to manage in the current session only.
ssmgui creates an icon for each array that it finds connected to each server whose hostname you added to the list (unless the icon already exists). ssmgui displays these icons in the array selection area of the Storage System Manager window (see Figure 2-3). If an array is connected to two servers, only one icon appears.
Figure 2-2 shows a Host Administration window with a server hostname.
If ssmgui cannot communicate with a server you added to the list, it displays an error message.
To remove a server hostname, select it in the Host Administration window and click Remove. The RAID GUI asks for confirmation.
Save the list as explained in “Using the Host Administration Window to Add Servers”. ssmgui removes the icon for that array unless the array is connected to another server on the list. It also removes the hostname from the Host Administration window.
Selecting an Array to ConfigureThis section explains how to use RAID GUI features to select an array to configure:
Using the Array Selection Filter (Managing and Unmanaging Arrays)Use the array selection filter (Arrays Accessible By field) near the top of the Storage System Manager window to display only those arrays on a server. Figure 2-3 points out this portion of the Storage System Manager window, and other features.
Other ways to specify arrays are as follows:
To display the icon for a specific array for all managed servers, follow these steps:
In the array selection area, select the icon for each array that you do not want to manage.
From the Array menu, choose Unmanage. ssmgui removes the icons for the selected arrays from the array selection area.
| Note: This display does not affect the list of managed servers and their arrays as listed in the Host Administration menu.
|
To display icons for all arrays for one managed server:
If the Arrays Accessible By field (see Figure 2-3) shows the hostname of the server whose arrays you want represented in the array selection area, do nothing.
If this field does not show the server hostname, pull down the selection list and select the desired server hostname. In the array selection area, ssmgui shows icons for arrays that it finds connected to this server.
To display icons for specific arrays for one managed server only, follow these steps:
If necessary, select the desired server hostname in the Arrays Accessible By field (see Figure 2-3). In the array selection area, ssmgui shows icons for arrays that it finds connected to this server.
In the array selection area, select the icon for each array that you do not want to manage.
From the Array menu, choose Unmanage. ssmgui removes the icons for the selected arrays from the array selection area.
To convert unmanaged arrays to managed arrays, follow these steps:
Choose Manage Arrays from the File menu. The Manage Arrays window opens, as shown in Figure 2-4.
If the Host Identifier field does not contain the name of the server with the unmanaged arrays that you want to start managing, pull down its selection list and select the server's hostname (or type it in the Host Identifier field). The names of the server's unmanaged arrays appear in the Unmanaged Arrays list on the left, as shown in Figure 2-4.
For each array you want to start managing:
Select the array name in the Unmanaged Arrays list.
Click the right arrow button. The array's name moves to the Selected Arrays list, on the right.
When the Selected Arrays list contains the names of only those unmanaged arrays that you want to start managing, click OK. An array icon for each selected array appears in the array selection area of the Storage System Manager window.
The array selection area (see Figure 2-3) contains an icon for each array specified by the array selection filter. The icon for an array consists of the array name and a graphic.
The default name for an array is determined by the host and device entries in the ssmagent.config file on the server connected to the array. You can change the array's default name by choosing Name in the Array menu or by using the name array button at the left end of the toolbar. Changing the name does not affect the configuration file.
Enclosure Health and AccessibilityThe color of an icon for an array indicates its health:
The graphic for an array icon indicates the status of the array:
Accessible: ssmgui can communicate with the array.
Inaccessible: ssmgui cannot communicate with the array because its name is wrong in the ssmagent configuration file on its server, or because the ssmagent on its server was started by a user who was not logged in as root.
The icon for an inaccessible array has a shield over it:
Unsupported: The device entry in the ssmagent configuration file on its server is for a device that ssmgui does not support (for example, an internal disk on the server).
The icon for an unsupported array has a question mark over it:
Accessible, but faulted: ssmgui can communicate with the array, but one or more components is faulty. Chapter 5, “Identifying and Correcting Failures,” explains how to investigate and correct faulted components.
The icon for a faulted array is amber and contains an F:
| Note: The external Fibre Channel Hub is transparent to ssmgui and the command-line interfaces, and thus does not appear in any RAID GUI windows or RAID CLI output.
|
Icons Connected to More Than One ServerOnly one icon for each array appears in the array selection area even it if is connected to more than one server.
To determine if an array is connected to more than one server, use the Array Information button or choose Information from the Array menu. This process is explained in “Using the Array Information Window” in Chapter 4.
Using Storage System Manager Toolbar ButtonsYou can use the toolbar in the Storage System Manager window for many steps explained in this guide. Figure 2-5 shows toolbar buttons and their functions.
These buttons duplicate functions available in the Storage System Manager window's Array menu.
 | Tip: When you select an array, click the right mouse button in the array selection area to bring up a menu with the same functions as the toolbar buttons.
|
In the array selection area, click the icon for the array. A black box appears around the icon, as shown in Figure 2-6.
To display the Array menu when you select an array, use the right mouse button and hold the button down. The black box appears around the icon and the Array menu drops down to the right of the icon.
To select more than one array at a time:
or
The black box appears around each icon you select.
Using the Array Configuration Window to Set Up the ArrayIn the Storage System Manager window, choose Configure from the Array menu. The Array Configuration window appears; Figure 2-7 shows important features.
The title bar identifies the array that is represented in the window.
Using Array Toolbar ButtonsThe Array Configuration window has two toolbars, the array toolbar and the LUN toolbar. Figure 2-8 shows array toolbar buttons and their functions. (LUN toolbar buttons are shown in Figure 3-1 in Chapter 3.)
These buttons duplicate functions available in the Array Configuration window's Array menu.
To bind RAID 3 LUNs or to use read or write caching for other types of LUNs, you must have the required hardware, and you must specify the array memory partitions before binding the LUNs. This section explains
SP memory is divided into three memory banks:
front-end memory: read and write memory banks (DIMMs) on the SP board (at least 16 MB in each)
The read and write caches are dedicated DIMMs on the storage processor (SP); DIMMs are available in 8-MB, 32-MB, and 128-MB sizes. With the maximum of four 128-MB DIMMs installed, the maximum for each type of cache is 512 MB.
back-end memory: available memory in the disk enclosures
| Note: For a Fibre Channel storage system, the front end consists of the host, the SPs, and the communication between them, and the back end consists of the disk modules.
|
CPU or control RAM: the licensed internal code (LIC) running on the SPs uses 1 MB (mixed mode disabled) or 8 MB (mixed mode enabled) from the read memory and 1 or 8 MB from the write memory on the SP board, plus a dedicated memory bank on the board. (Mixed mode is explained later in this chapter, in “Bandwidth Mode and Mixed Mode”.)
LUNs other than RAID 3 use read cache and write cache; RAID 3 LUNs use RAID 3 memory. The LIC allocates the read cache partition to the back-end memory, the write cache partition to front-end memory, and the RAID 3 partition across both. The LIC uses the control RAM to manage the read and write caches and the RAID 3 partition (for example, it stores the cache page tables in the control RAM).
Memory that you can allocate to the read and write caches or the RAID 3 partition is the unallocated memory in the front-end and back-end memory; it is known as user free memory.
Size, allocation, and availability ranges for RAID 3 memory differ from those for the read cache and write cache. RAID 3 LUNs cannot be bound in the same array as other LUN types (except hot spares).
You can enable or disable read cache for either SP in an array. The read cache memory that you allocate when you partition memory is shared by all the SP's LUNs for which the read cache is enabled. Read cache is independent on each SP.
Unlike read cache, write cache is always the same (mirrored) on both SPs in an array. You can enable and disable an SP's write cache or change its size; all such changes happen automatically for the array's other SP as well. The write cache memory is shared by all LUNs for which the write cache is enabled.
Besides the array caches, you can assign a specific cache to a particular LUN. However, LUN caching is not enabled until array caching is enabled.
Hardware Requirements for Caching: LUNs Other than RAID 3Hardware requirements for caching are as follows:
Necessary hardware is installed by a qualified Silicon Graphics System Support Engineer; contact your authorized service provider.
Hardware Requirements for RAID 3 LUNsFor RAID 3 LUNs, each SP must have at least 16 MB each for read and write memory locations; if 8-MB DIMMs are used, each read memory and each write memory has 16 MB; with 32-MB DIMMs, each read memory and each write memory has 32 MB. As mentioned before, the system allocates the RAID 3 memory across read and write DIMMs.
Memory, Bandwidth Mode, and Mixed ModeTo optimize performance for RAID 3 LUNs and other LUN types, the storage management software provides two modes, called bandwidth and mixed. Bandwidth mode optimizes performance for RAID 3 transfers; mixed mode, the default, is for other types of LUNs.
The software toggles between these two modes; enabling bandwidth mode is the same as disabling mixed mode. These modes are explained fully at “Bandwidth Mode and Mixed Mode”.
Memory availability changes dynamically depending on the type of LUNs bound and on whether bandwidth or mixed mode is set. For releases starting in the third quarter of calendar 1998, the following applies:
With mixed mode enabled (as shipped), read cache, write cache, and RAID 3 memory are factory-set to 0. If RAID 3 LUNs are configured, the software automatically sets RAID 3 memory to 2 MB on each SP, which is the minimum required for accessing RAID 3 LUNs.
With bandwidth mode enabled (optimization for RAID 3 LUNs), read and write cache are each set to 0 and RAID 3 memory is factory-set to 14 MB. For RAID 3 LUNs, you can leave this setting, which is the optimum and recommended level, or you can change it to any amount between 2 MB and 14 MB.
Memory Partitioning Points Note the following points about allocating memory partitions to specific memory areas:
Changing the RAID 3 partition size causes the LIC to reboot the array. Rebooting restarts the SPs in the array, terminating all outstanding I/O to the array. If you plan to change RAID 3 partition size, make sure that no users are conducting I/O with any filesystems or partitions on the array, and unmount these filesystems or partitions. After the process is completed, remount the filesystems or partitions and notify the users.
You might not be able to allocate all of the user free memory to a specific partition.
For example, because the read cache partition is only in the back-end memory (disk storage), you cannot allocate user free memory in the front end to the read cache partition. Similarly, because the write cache partition is only in the front end, you cannot allocate any user free memory in the back end to the write cache partition.
You might not be able to increase the size of one partition without decreasing the size of another partition.
When you increase the size of a partition, ssmgui tries to assign user free memory to the partition you are increasing. When ssmgui allocates all the user free memory that is available for the partition, it attempts to take memory from another partition to accommodate your request. As a result, when you increase the size of one partition, the size of another partition might decrease. This size change is shown graphically in the Memory Partition window, as discussed on page 39 in “Partitioning Array Memory.”
For example, the size of the RAID 3 partition influences the size of the read cache partition, so that when you increase the size of the read cache partition, the size of the RAID 3 partition might decrease.
Note that when you decrease the size of a partition, the amount of deallocated memory returns to the user free partition, and the sizes of the other partitions do not increase.
You might not be able to allocate all user free memory to a partition if the page tables for the new partition sizes would not leave enough free control RAM for the LIC.
The size of the page tables for the read or write cache partition are affected by the size of the cache page size in addition to the size of the partition. A small cache page size results in more cache pages, and thus a larger page table and less free control RAM for the LIC. Conversely, a large cache page size results in fewer cache pages, and therefore smaller cache page tables and more free control RAM for the LIC. (You can set the cache page size after binding the non-RAID 3 LUNs; this procedure is explained in “Changing Cache Page Size” in Chapter 3.)
| Note: Binding other types of LUNs with RAID 3 LUNs in the same array is not supported.
|
Partitioning Array Memory You partition memory for each SP using the Memory Partition window, which depicts the enclosure's two SPs and the amount of memory allocated to various functions.
 | Caution: If you change the RAID 3 partition size, ssmgui reboots the array. Rebooting restarts the SPs in the array, terminating all outstanding I/O to the array. If you plan to change RAID 3 partition size for existing RAID 3 LUNs, make sure no users are conducting I/O with any filesystems or partitions on the array, and unmount these filesystems or partitions. After the process is completed, remount the filesystems or partitions and notify the users.
|
To allocate memory, follow these steps:
In the Storage System Manager window, select the arrays whose SP memory you want to partition: click the configure array button near the left end of the array toolbar, or choose Configure from the Array menu.
To specify memory for one array, double-click the array icon to open the Array Configuration window.
If desired, and if the array has the necessary hardware as explained in “Hardware Requirements for Caching: LUNs Other than RAID 3”, you can change RAID 3 memory size, or change read cache and write cache size for other types of LUNs:
To set RAID 3 partition size:
Make sure bandwidth mode (optimization for RAID 3 LUNs) is enabled (mixed mode is disabled): In the Array menu of the Array Configuration window, see that the menu choice reads Enable Mixed Mode. If the item reads Enable BandWidth Mode, as in Figure 2-9, click on this menu choice to enable bandwidth mode.
The system is shipped with mixed mode enabled.
In the Array Configuration window, click the partition memory button, or choose Partition Memory from the Array menu. The Memory Partitions window for the array opens, as shown in Figure 2-10.
Below the pie charts, the fields—Read Cache, Write Cache, RAID 3, Extended, LIC System, and User Free—show the amount of SP memory allocated to those partitions of SP memory.
User free memory is the amount of memory not already allocated to the other memory partitions; its default size is equal to the total memory minus that required for the LIC. Thus, the User Free field shows the amount of memory you can allocate to the other partitions. See “Memory Partitioning Points” for an explanation of the interaction and requirements of memory partitions.
The memory indicated for LIC is the sum of the amount taken by the LIC (8 MB from each cache with mixed mode enabled, 1 MB with it disabled) plus the CPU RAM required by the SP; depending on the revision level, this amount is 8 MB or 16 MB.
| Note: Extended memory is always 0 and cannot be changed.
|
In the Memory Partition window, drag the RAID 3 slider until the desired number of MB appears in the field to the right of the slider, or type the size directly in the field.
| Caution: If you change the RAID 3 partition size, the software reboots the array. Rebooting restarts the SPs in the array, terminating all outstanding I/O to the array. Before you change RAID 3 partition size for existing LUNs, make sure no users are conducting I/O with any filesystems or partitions on the array, and unmount these filesystems or partitions.
|
As you move the RAID 3 slider, the pie charts for both SPs change to show the portion of memory allocated to each SP's RAID 3 partition, because it must be the same size on each SP.
| Caution: If valid RAID 3 LUNs have already been bound and you reduce RAID 3 memory to 0 MB, the RAID 3 LUNs are unowned and are no longer accessible after a reboot.
|
To set read and write cache partition size for LUNs other than RAID 3:
For SP A read cache partition, drag the Read Cache slider until the desired number of megabytes appears in the field to the right of the slider, or by typing the size directly in the field. As you move the slider, the pie chart shows the amount of memory allocated to read cache. (Memory partitions are color-coded; the legend near the bottom of the window explains color coding.)
| Note: Array read caching for SP A is disabled if the SP's read cache partition is 0 MB. It stays disabled you until allocate memory to the partition, and enable caching as explained in “Partitioning Array Memory”.
|
Follow the same procedure to set SP B read cache partition size. Read caches for the two SPs are independent of each other.
Set SP A write cache partition size by dragging the Write Cache slider until the desired number of megabytes appears in the field to the right of the slider, or by typing the size directly in the field.
As you move the slider, the pie charts for both SPs change to show the portion of memory allocated to each SP's write cache partition, because the write cache partition must be the same size on each SP (unlike the read cache partition). The slider and number field for write cache on the other SP also reflect the changing memory setting.
| Note: Array write caching for SP A is disabled if the SP's write cache partition is 0 MB. It stays disabled you until allocate memory to the partition. Use a minimum of 2 MB for the write cache partition on each SP.
|
To put the memory allocation changes into effect, click OK at the lower left of the Memory Partition window.
If you have LUNs other than RAID 3 and if read and write cache have values other than 0 MB, a confirmation window appears, informing you that read and write cache are disabled for this operation, and reenabled after it is complete; see Figure 2-11.
Click Yes in the confirmation window to reallocate memory.
If you are changing RAID 3 memory size, the reboot starts.
If caching has not been enabled and you have allocated memory to the Read Cache or Write Cache partition, use the Array Configuration window to enable caching:
To enable array write caching for both SPs, click the write cache enable button (at the far left of the array toolbar), or choose Array > Write Cache State > Enable.
To enable read caching for SP A, click the SP A enable read cache button or choose Array > Read Cache State >SP A > Enable.
To enable read caching for SP B, click the SP B enable read cache button or choose Array > Read Cache State >SP B > Enable.
| Note: After you enable read or write caching, the software informs you that enabling was successful, regardless of whether the array has the necessary hardware for such caching.
|
You may have to restart the agent on the server connected to the array whose memory you just partitioned; see next section, “Planning the Bind.”
If you have unmounted filesystems or partitions in the array to change RAID 3 partition size, remount the filesystems or partitions and notify the users.
You have set up the array(s) to perform read caching, write caching, or both, or you have set up RAID 3 memory. The array uses the default values for caching or memory parameters. To change these settings, see Chapter 3, “Reconfiguring and Fine-Tuning.” Otherwise, proceed to the next section, “Planning the Bind.”
You must bind disk modules into LUNs so that the server's operating system can recognize them. Consider the following before you bind LUNs:
number of disk modules and their capacities
parameters for the LUN, such as SP ownership, caching, and so on
possible effect of bind operation on operations in progress (whether rebooting is required)
Table 2-1 summarizes the number of disk modules you can have in LUNs, and other important factors for binding.
Table 2-1. Binding Disk Modules
LUN Type
| Disk Modules
| Notes
|
|---|
Any LUN
| Unbound disk modules only;
all must have the same
capacity
| For LUNs except RAID 3, you can deallocate RAID
3 memory (see “Setting Up Array Memory”). Binding RAID 3 LUNs with other LUN
types in the same array is not supported.
| RAID 5
| 3 minimum
(5 recommended minimum),
16 maximum
| You can bind one less module per LUN that you
eventually use by selecting an empty slot icon.
However, the LUN operates in a degraded mode
until a disk module is installed in the empty slot
and the array integrates it into the LUN. You can
select the modules in any order.
| RAID 3
| 5 or 9
| You can bind one less module per LUN than you
eventually use by selecting an empty slot icon.
However, the LUN operates in a degraded mode
until a disk module is installed in the empty slot
and the array integrates it into the LUN. You can
select the modules in any order.
Check RAID 3 memory before binding, as
explained in “Partitioning Array Memory”. If you do not allocate adequate memory to
the RAID 3 partition, the LUN is unowned.
Binding other LUN types in the same array is not
supported.
| RAID 1
| 2
|
| RAID 1/0
| Even number:
4 minimum, 16 maximum
| Disk modules are paired into mirrored images in
the order in which you select them: see Figure 2-14
on page 52.
| RAID 0
| 3 minimum, 16 maximum
| You can select the modules in any order.
| Individual
disk
| 1
|
| Hot spare
| 1
| You cannot bind as hot spares disk modules 0:0
through 0:8.
The capacity of a hot spare must be at least as great
as the capacity of the largest disk module that it
might replace.
For a RAID 3 LUN, only one hot spare is used in
case of disk failure; a second hot spare is not used
in case of a second disk failure.
|
This section contains the following topics:
| Note: Binding takes a while; how long varies with the type of SP and size of the disk modules. A RAID 3 or RAID 5 LUN might take more than four hours.
|
Bandwidth Mode and Mixed Mode Using a menu choice in the Array menu of the Array Configuration window, you must enable bandwidth mode (the default) for the LUNs in an array or enable mixed mode, depending on LUN type. The software toggles between these two modes; enabling bandwidth mode is the same as disabling mixed mode.
The modes affect the operation of different RAID types in different ways, as shown in Table 2-2.
Table 2-2. Bandwidth Mode and Mixed Mode
If Menu Item Is...
| Then Mode Is...
| RAID 5, 1, 1/0, 0
| RAID 3 and Hot Spares
|
|---|
Enable BandWidth Mode
| Mixed
| These LUNs can be bound and
accessed.
| RAID 3 LUNs are not accessible.
| Enable Mixed Mode
| Bandwidth
| These LUNs can be bound, but not
accessed until mixed mode is enabled.
| These LUNs can be bound and accessed;
RAID 3 operation is optimized.
|
Note the following:
If you are binding or accessing RAID 3 LUNs, enable bandwidth mode, which optimizes performance for the array's RAID 3 transfers. With bandwidth mode enabled (the menu item in the Array menu of the Array Configuration window reads Enable Mixed Mode), you can bind other RAID types, but you cannot access them. The system is shipped with bandwidth mode enabled.
If you are binding or accessing RAID 5, 1, 1/0, or 0 LUNs, enable mixed mode (the menu item must read Enable BandWidth Mode).
If RAID 5, 1, 1/0, 0 LUNs are bound when bandwidth mode is enabled, they are unowned; if you then enable mixed mode, they become owned.
If mixed mode is enabled, disabling it (that is, clicking Enable BandWidth Mode in the Array menu of the Array Configuration window) reboots the array's SPs and restarts the LIC. Before beginning this process, make sure that no users are conducting I/O with any filesystems or partitions on the array.
RAID 3 and hot spares work whether bandwidth mode or mixed mode is enabled, but RAID 3 is optimized if bandwidth mode is enabled.
Before enabling mixed mode, unbind all RAID 3 LUNs and set RAID 3 memory to 0. Before enabling bandwidth mode, set the write cache partition to 0 and disable read and write cache.
Steps for these processes are included in the instructions for binding LUNs later in this chapter.
Arrays set up according to instructions in “Setting Up Array Memory” use default values for low and high watermarks and for read caching prefetch parameters. To change these settings, follow instructions in “Binding Disk Modules”.
When you set up a LUN, it uses certain bind parameters:
| Rebuild priority | |
Priority that the array assigns for reconstructing the data on either a hot spare or a new disk module that replaces a failed disk module in a LUN: ASAP (as soon as possible; 10 to 15 minutes per GB), High (15 to 30 minutes per GB), Medium (45 to 60 minutes per GB), or Low (75 to 90 minutes per GB). The rebuild priority applies to all RAID LUNs except RAID 0, individual disk LUNs, and hot spares. The default priority is High.
| | Verify priority | | Priority that the array allots for checking parity: ASAP (as soon as possible), High, Medium, or Low. If an SP detects parity inconsistencies, it starts a background process to check all the parity sectors in the LUN. The priority you specify determines the amount of resource the SP devotes to verifying instead of to normal I/O activity.
| | Element size (stripe element size) | |
Number of disk sectors that the array can read or write to a single disk module without requiring access to another disk module (assuming that the transfer starts at the first sector in the stripe). The stripe element size can affect the performance of a RAID 5 or RAID 1/0 LUN. A RAID 3 LUN has a fixed stripe element size of one sector. The smaller the stripe element size, the more efficient the distribution of data read or written. However, if the stripe size is too small for a single I/O operation, the operation requires access to two stripes, which causes the hardware to read and/or write from two disk modules instead of one.
| | Default SP | | The SP that assumes ownership of the LUN after the array's power is turned off and then on again. If the array has two SPs, you can choose to bind some LUNs using one SP as the default, and the rest using the other SP as the default. This tactic balances the load across the SPs or, in a dual-server configuration, establishes the primary route for the LUN. The primary route to a LUN is the route through the default SP; the secondary route is through the other SP.
| | Read cache | | You can enable or disable read cache for either LUN. The read cache memory that you allocate when you partition memory is shared by all the SP's LUNs for which the read cache is enabled. Read cache is independent on each SP.
| | Write cache | | You can enable or disable write cache for either LUN. The write cache memory that you allocated when you partitioned memory is shared by all LUNs for which the write cache is enabled.
| | Auto assignment state | |
Auto assignment state controls the ownership of the LUN when one SP fails in a array with two SPs. With auto assign enabled, if the SP that owns a LUN fails and the server tries to access that LUN through the second SP, the second SP assumes ownership of the LUN so the access can occur. The second SP continues to own the LUN until the SP's power is turned off and on again, at which point ownership of each LUN returns to its default SP. If auto assign is disabled in an array with two SPs, the other SP does not assume ownership of the LUN, so the access to the LUN does not occur. It is recommended that auto assign be left disabled, which is the default for all LUNs (except hot spares, to which it does not apply).
| | Minimal latency reads | |
For RAID 3 LUNs only, minimal latency reads provide a more constant bandwidth between the SP and the server when the bandwidth between the disk modules in a RAID 3 LUN and the SP decreases because of a slow response from a disk module. Minimal latency reads change how the SP responds to a request from the server to read a RAID 3 LUN.
When minimal latency reads are disabled, the SP responds to a read request to a RAID 3 LUN with n disk modules by issuing a read request to the n-1 disk modules in the RAID 3 LUN that contain data. It performs a checksum on the data as it receives the data from the n-1 disk modules. If the checksum is correct, it sends the data to the server; if the checksum is incorrect, it reads the parity from the remaining disk module and uses it to reconstruct the data before sending the data to the server.
When minimal latency reads are enabled, the SP responds to a read request to a RAID 3 LUN with n disk modules by issuing a read request to all n disk modules in the LUN. It performs a checksum on the data as it receives it from the first n-1 drives to respond. If the checksum is correct, it sends the data to the server; if the checksum is incorrect, it uses the parity it has received to reconstruct the data before sending the data to the server.
Because minimal latency reads always require access to all the disk modules in the RAID 3 LUN, they nominally decrease the bandwidth between the disk modules and the SPs by about 25%; for a RAID 3 LUN with nine disk modules, it decreases by about 12%.
|
Not all of these parameters apply to all RAID types, as summarized in Table 2-3.
Table 2-3. RAID Types and Bind Parameters
RAID Type
| Default SP
| Rebuild Time
| Stripe (Element) Size
| Verify Priority
| Auto Assign
| Caching
|
|---|
RAID 5
| Required
| Required
| Required
| Yes
| Disable
| Read and write
| RAID 3
| Required
| Required
| Not required
| Yes
| Disable
| Not required
| RAID 1
| Required
| Required
| Not required
| Yes
| Disable
| Read and write
| RAID 1/0
| Required
| Required
| Required
| Yes
| Disable
| Read and write
| RAID 0
| Required
| Not required
| Required
| N/A
| Disable
| Read, write, or both
| Individual disk
| Required
| Not required
| Not required
| N/A
| Disable
| Read, write, or both
| Hot spare
| Not required
| Not required
| Not required
| N/A
| Not required
| Not required
|
The LUN number must be specified for all RAID types in Table 2-3. RAID 3 also requires minimal latency reads.
Table 2-4 summarizes the values that ssmgui sets for standard LUN parameters.
Table 2-4. Default LUN Parameters
RAID Type
| Rebuild
Priority
| Verify
Priority
| Element
Size
|
Disk Modules
| Read and Write
Caching
|
Auto Assign
| Minimal Latency
Reads
|
|---|
RAID 5
| High
| High
| 128
| 5
| Enabled
| Disabled
| N/A
| RAID 3
| High
| High
| N/A
| 5
| N/A
| Disabled
| Disabled
| RAID 1
| High
| High
| N/A
| 2
| Enabled
| Disabled
| N/A
| RAID 1/0
| High
| High
| 128
| 6
| Enabled
| Disabled
| N/A
| RAID 0
| N/A
| N/A
| 128
| 5
| Enabled
| Disabled
| N/A
| Individual disk
| N/A
| N/A
| N/A
| 1
| Enabled
| Disabled
| N/A
|
For all LUN types, the default SP is determined by load balancing; if only one SP is connected, that SP is the default.
The bind procedure varies depending on the type of LUN you are binding. This section contains the following topics:
| Note: Using LUN buttons in the Storage System Manager toolbar (see Figure 2-5) to bind LUNs is not recommended. Use them only under the following conditions:
|
You are not changing the mode setting (bandwidth or mixed); see “Bandwidth Mode and Mixed Mode”.
You are not changing any bind parameters from the defaults; see Table 2-4.
There are no performance considerations; disk modules are bound in order of availability and cannot be selected or specified.
You are using the standard number of disk modules for the LUN; see Table 2-4.
Setting Bind Parameters: RAID 5, 1, 1/0, or 0To set bind parameters for RAID 5, 1, 1/0, or 0, follow these steps:
Because you must enable mixed mode after you bind these types of LUNS, make sure that no users are conducting I/O with any filesystems or partitions on the array. If necessary, notify the users and unmount these filesystems or partitions.
If you plan to use caching, allocate memory as necessary; see “Setting Up Array Memory”. You can deallocate any memory allocated to RAID 3 memory (binding RAID 3 memory and other types of LUNs in the same array is not supported). This process may involve rebooting.
To speed up binding, turn off automatic polling if it is on (see the field at the lower right of the Storage System Manager window): click the button at the far right of the Storage System Manager toolbar (as shown in Figure 2-8).
To see status information during binding, poll for it manually by clicking the manual poll button next to the automatic poll button in the toolbar, or choose Poll from the Array menu. To see the polling information, double-click the LUN icon in the Array Configuration window to open the LUN Information window (see “LUN Configuration Information” in Chapter 4 for an explanation of fields in the window).
In the Storage System Manager window, select the array whose disk modules you want to bind. You can select more than one array if you want to bind all their disk modules into the same type of LUN.
Double-click on the array to open its Array Configuration window.
In the Array menu of the Array Configuration window, make sure mixed mode is enabled. If you see the menu choice Enable BandWidth Mode, then mixed mode is currently enabled; proceed to step 7.
If you see the menu choice Enable Mixed Mode, then mixed mode is currently disabled (bandwidth mode is enabled instead), and you must enable it by selecting this choice.
Open the Bind LUNs dialog for the array(s): click the button near the middle of the LUN (lower) toolbar, or choose Bind LUN from the Array menu. Figure 2-12 shows an example of the Bind LUNs dialog.
Select disk modules for the LUN. If the array has disk modules in more than one enclosure, select the enclosures containing the disk modules you want to bind:
To select disk modules from all enclosures in the array: If the Unbound Disks field contains All Chassis, continue to step 9. If it contains the name of an enclosure, pull down its selection list and choose All Chassis.
To select disk modules from one enclosure in the array: If the Unbound Disks field shows the name of the enclosure you want, proceed to step 9. Otherwise, choose the enclosure name from the pull-down menu.
In the Unbound Disks area, select the disk modules that you want to bind into a LUN, and click the right arrow button. (Alternatively, you can use both mouse buttons to drag and drop the disk modules from the Unbound Disk area to the Bind Disks area.) Figure 2-13 shows disk modules selected for binding.
See Table 2-1 on page page 43 for the number of disks in various types of LUNs.
| Tip: To select multiple consecutive disk modules, click the first disk module icon and drag the cursor over the other disk modules. (Alternatively, you can hold down the Ctrl key and click disk modules.)
|
If you move a wrong disk module to the Bind Disks area, select it and click the left arrow button to move it back to the Unbound Disks area. (Or drag and drop it into the Unbound Disks area.)
If you are binding a RAID 1/0 LUN, the order in which you select modules is important; Figure 2-14 diagrams this order.
In the RAID Type field, select the RAID type. The list displays only RAID types that are available for the number of disk modules you selected.
Make sure Auto Assign in the Options section of the Bind LUNs dialog is not selected (the default), as shown in Figure 2-15. For more information on auto assign, see page 47.
To change other bind parameters in the Bind LUNs dialog:
LUN ID: Select another LUN hexadecimal identifier (ID) from the LUN ID field.
The default LUN ID is the next hex number available, starting with 0 and ending with 1f. The list displays only numbers that are available. The default number is 0 for the first LUN that you bind, regardless of the number of SPs or servers attached to the array. The default number for the second LUN you bind is 1; for the third LUN, it is 2; for the fourth LUN, it is 3, and so on. You can specify a nondefault number if desired. After you bind a LUN with a nondefault number, the default number for the next LUN is the lowest number you skipped. The maximum number of LUNs is 32.
For hot spares, assign LUN numbers starting with the highest number available and continue downwards.
Rebuild Priority: If the LUN is not a RAID 0 LUN, individual disk, or hot spare, select the priority: ASAP, High, Medium, or Low. For more information on rebuild priority, see page 46.
Verify Priority: If the LUN is not a RAID 0 LUN, individual disk, or hot spare, select the priority: ASAP, High, Medium, or Low. For more information on verify priority, see page 46.
Element Size: For a RAID 0, RAID 1/0, or RAID 5 LUN, if you want the LUN to have a stripe element size with a different number of sectors from the number in the Element Size field, select the desired number of sectors from the field list.
Generally, use the smallest stripe element size that rarely forces access to another stripe. The default stripe element size for RAID 5 is 128 sectors. Any size you choose should be an even multiple of 16 sectors; supported values are 4, 8, 16, 32, 64, and 128. For more information on element size, see page 46.
Read Cache: For LUNs other than RAID 3 or hot spares, to set the state of the read or write cache for the LUN, click the SP's read cache button near the left end of the array toolbar. The display indicates whether read or write cache for the default SP is enabled or disabled for the LUN, as shown in Figure 2-16.
Enabling read cache is recommended for any type of LUN (except a RAID 3 LUN or hot spare).
The read cache memory that you allocated when you partitioned memory is shared by all LUNs for which the read cache is enabled.
Write Cache: For LUNs other than RAID 3 or hot spares, to set the state of the read or write cache for the LUN, click the SP's read cache button near the left end of the array toolbar. The display indicates whether write cache for the default SP is enabled or disabled for the LUN.
Enabling write cache for a RAID 5 LUN is highly recommended; enabling it for other LUN types for which it is possible is also recommended. Write cache is always mirrored and thus requires two SPs.
The write cache memory that you allocated when you partitioned memory is shared by all LUNs for which the write cache is enabled.
Default SP: If the LUN is not a hot spare, click the button for the other SP to change the LUN's default owner to that SP. This option is available only for arrays with two SPs. For more information on default SP, see page 46.
When all bind parameters for the LUN are the way you want them, click Bind at the lower left in the Bind LUNs dialog.
In the confirmation window that opens, click Yes to start the bind operation. A window opens stating that the bind operation was successfully initiated; click OK.
A blue icon for the LUN appears in the Unowned LUNs area in the Array Configuration window. A small letter T in the icon indicates its transitional state.
Binding takes a while; how long varies with the type of SP and size of the disk modules. (A RAID 5 LUN might take more than four hours.) When polling determines that the bind operation is completed, the LUN icon moves to the selection area for its default SP and becomes gray.
| Note: To stop a bind that is in progress, use the RAID CLI unbind subcommand; see “unbind” in Chapter 6. For agent revision 2.04.xx and later, use the unbind button in the LUN toolbar of the Array Configuration window; see Figure 3-1 on page 66 in Chapter 3.
|
Once the LUN is bound, change the mode setting from bandwidth to mixed: choose Enable Mixed Mode from the Array menu. (“Enable Mixed Mode” means that mixed mode is currently disabled and bandwidth mode is enabled instead.)
A confirmation window opens, warning you that the system will reboot; see Figure 2-17.
Click Yes to enable mixed mode.
After the reboot, remount any filesystems or partitions you unmounted at the beginning of the bind.
Setting Bind Parameters: RAID 3 and Hot SparesTo set bind parameters for RAID 3 and hot spares, follow these steps:
Make sure the memory allocated for RAID 3 is as you want it; see “Setting Up Array Memory”. This process may involve rebooting.
To speed up binding, turn off automatic polling if it is on; see step 3 on page 49.
In the Storage System Manager window, select the array whose disk modules you want to bind. You can select more than one array if you want to bind all their disk modules into the same type of LUN.
Double-click on the array to open its Array Configuration window.
In the Array menu of the Array Configuration window, look for the mixed mode menu choice:
If you see Enable Mixed Mode, then mixed mode is currently disabled (bandwidth mode is enabled); proceed to step 6.
If you see Enable BandWidth Mode, then mixed mode is currently enabled; you must disable it.
If you must disable mixed mode, follow these steps:
| Caution: Disabling mixed mode reboots the array's SPs and restarts the LIC.
|
Make sure that no users are conducting I/O with any filesystems or partitions on the array. If necessary, notify the users and unmount these filesystems or partitions.
In the Memory Partition window (see page 39), set write cache to 0.
Disable read and write cache: click the read cache and write cache disable buttons in the array toolbar (see Figure 2-8 on page 32), or choose Array > Write Cache State > Disable for each SP, and Array > Read Cache State > Disable.
Click Enable BandWidth Mode in the Array menu of the Array Configuration window. A confirmation window opens, as shown in Figure 2-18.
If cache is enabled, the software disables it for the operation and then restores it; Figure 2-19 shows the warning message in this case.
Click Yes in the warning window.
Select all the disk modules for the LUN; for RAID 3, select either 5 or 9. See instructions at step 7 on page 50. In binding hot spares, you can select as many disks in as many enclosures as you like, and bind them all at the same time.
| Note: Binding other types of LUNs with RAID 3 LUNs in the same array is not supported (except hot spares).
|
In the RAID type field, select RAID 3 or Hot Spare Only RAID types that are available for the number of disk modules you selected appear in the list.
For RAID 3 LUNs, make sure Auto Assign in the Options section of the Bind LUNs dialog is not selected (the default), as shown in Figure 2-16. For more information on auto assign, see page 47.
Change other bind parameters:
LUN ID: see page 53.
For hot spares, assign LUN numbers starting with the highest number available and continue downwards.
Rebuild Priority and Verify Priority: For a RAID 3 LUN, if you want a different rebuild or verify priority from that in the field, select it from the pull-down list. For more information on these options, see page 46.
Read Cache, Write Cache: Enabling these is not supported for RAID 3 LUNs or hot spares.
Minimal Latency Reads: To change this option, which is available only for RAID 3 LUNs, click Minimal Latency Reads in the Options section. For more information on minimal latency reads, see page 47.
Default SP: If the LUN is not a hot spare, click the button for the other SP to change the LUN's default owner to that SP. This option is available only for arrays with two SPs. For more information on default SP, see page 46.
If your application that is writing to the RAID 3 LUN is single-threaded and performance is more important than data integrity, you can increase the performance of these applications by enabling RAID 3 buffering:
Display the Array Configuration window for the array whose RAID 3 write buffering you want to enable.
Determine whether the array's RAID 3 write buffering is enabled: click the array information window button near the right end of the array toolbar in the Array Configuration window, or choose Array Information from the Array menu. A window opens; Figure 2-20 shows an example.
If the array's RAID 3 write buffering is not already enabled, choose Enable RAID3 Write Buffering from the Array menu.
In the confirmation window that opens, click Yes.
(To disable RAID 3 write buffering, choose Disable RAID3 Write Buffering from the Array menu, and click Yes in the confirmation window.)
When all bind parameters for the LUN are the way you want them, click Bind at the lower left in the Bind LUNs dialog.
In the confirmation window that opens, click Yes to start the bind operation. A window opens stating that the bind operation was successfully initiated; click OK.
A blue icon for the LUN appears in the Unowned LUNs area in the Array Configuration window. A small letter T in the icon indicates its transitional state.
Binding takes a while; how long varies with the type of SP and size of the disk modules. (A RAID 3 LUN might take more than four hours.) When polling has determined that the bind operation is completed, the LUN icon moves to the selection area for its default SP and becomes gray.
| Note: To stop a bind that is in progress, use the RAID CLI unbind subcommand; see “unbind” in Chapter 6. For agent revision 2.04.xx and later, use the unbind button in the LUN toolbar of the Array Configuration window; see Figure 3-1 on page 66 in Chapter 3.
|
Once the LUN is bound, change the mode setting from mixed to bandwidth:
| Caution: Enabling bandwidth mode (disabling mixed mode) reboots the array's SPs and restarts the LIC.
|
Make sure that no users are conducting I/O with any filesystems or partitions on the array.
Choose Enable BandWidth Mode from the Array menu of the Array Configuration window. A confirmation window opens, as shown in Figure 2-17.
In the confirmation window that appears, click Yes.
If you have unmounted filesystems or partitions to disable mixed mode, remount the filesystems or partitions and notify the users.
| Note: A RAID 3 LUN can use only one hot spare in case of disk failure. If a second hot spare is available and a second disk module in the RAID 3 LUN fails, the LUN does not use the second hot spare.
|
Making LUNs Available to the Server Operating SystemWhen all LUNs are assigned, make the LUNs available to the server's operation system:
or
Enter
scsiha -lp controllernumber; ioconfig -f /hw
|
For example:
scsiha -lp 3; ioconfig -f /hw
|
Enabling Command-Tagged QueuingCommand-tagged queueing (CTQ) allows multiple outstanding commands to a single SCSI target (that is, a LUN in a storage system), resulting in increased I/O performance.
The SP supports SCSI-2 queuing of requests for its LUNs. The requests are first-come, first-served in that all requests can be sent to one LUN from one initiator and cause a Queue Full status (unexpected SCSI status byte 0x28) to be returned for all other I_T_L SCSI selections. This condition continues until one of the outstanding requests completes and thus frees queue space. The SP can handle up to 250 CTQs.
If Queue Full status is returned for a given I/O request, that request is retried. If the requests cannot be sent to the SP after four retries, the request is aborted, which, in the case of a write request, can have unfortunate consequences.
When CTQ is enabled (with fx) for a given LUN, the default CTQ depth for the LUN is 2; this value is stored in the LUN's volume header. You must use fx to change this value. Table 2-5 defines the maximum CTQ depth values per LUN for single-hosted SPs and dual-hosted SPs.
Table 2-5. Maximum CTQ Depths per LUN
Number of LUNs
| Single-Hosted SPs
| Dual-Hosted SPs
|
|---|
1
| 250
| 125
| 2
| 125
| 62
| 3
| 83
| 41
| 4
| 62
| 31
|
For optimum system performance, enable command-tagged queuing. Table 2-6 shows performance benefits of CTQ.
Table 2-6. CTQ Performance Benefits for 2 KB Random Read, 16 Threads
CTQ
| avque
| r+w/s
| blks/s
| w/s
| wblks/s
| avwait (μs)
| avserv (ms)
|
|---|
Disabled
| 16.0
| 69
| 137
| 0
| 0
| 217.6
| 14.5
| Enabled
| 16.0
| 304
| 607
| 0
| 0
| 48.8
| 3.3
|
The fx program syntax is as follows.
fx -x “controllertype(controller_number,drive_number,lun_number)”
|
Follow these steps:
Enter the fx command with appropriate parameters; for example:
Output such as the following appears:
fx version 6.4, Aug 3, 1998
...opening dksc(6,2,2)
...controller test...OK
Scsi drive type == SGI RAID 5 0757
fx: Warning: bad sgilabel on disk
creating new sgilabel
----- please choose one (? for help, .. to quit this menu)-----
[exi]t [d]ebug/ [l]abel/ [a]uto
[b]adblock/ [exe]rcise/ [r]epartition/ [f]ormat
|
Update parameters; at the fx> prompt, enter
Output such as the following appears:
fx/label/set/parameters: Error correction = (enabled)
fx/label/set/parameters: Data transfer on error = (enabled)
fx/label/set/parameters: Report recovered errors = (enabled)
fx/label/set/parameters: Delay for error recovery = (enabled)
fx/label/set/parameters: Err retry count = (0)
fx/label/set/parameters: Transfer of bad data blocks = (enabled)
fx/label/set/parameters: Auto bad block reallocation (write) = (enabled)
fx/label/set/parameters: Auto bad block reallocation (read) = (enabled)
fx/label/set/parameters: Read ahead caching = (disabled)
|
At the Enable CTQ prompt, enter enable:
fx/label/set/parameters: Enable CTQ = (disabled) enable
|
At the CTQ depth prompt, enter 10:
fx/label/set/parameters: CTQ depth = (2) 10
|
Output such as the following appears.
fx/label/set/parameters: Read buffer ratio = (0/256)
fx/label/set/parameters: Write buffer ratio = (0/256)
* * * * * W A R N I N G * * * * *
|
At the following prompt in the last line above, enter yes:
about to modify drive parameters on disk dksc(6,2,2)! ok? yes
|
The following output appears:
----- please choose one (? for help, .. to quit this menu)-----
[exi]t [d]ebug/ [l]abel/ [a]uto
[b]adblock/ [exe]rcise/ [r]epartition/ [f]ormat
|
Type exit to exit fx. The following message appears:
label info has changed for disk dksc(6,2,2). write out changes? (yes)
|
Type y to write the changes to disk.
Origin FibreVault and Fibre Channel RAID Administrator's Guide
(document number: 007-3715-002 / published: 1998-12-23)
table of contents | additional info | download
Front Matter
About This Guide
Chapter 1. Fibre Channel RAID Graphical User Interface Features
Chapter 2. Using the RAID GUI to Configure Arrays
Chapter 3. Reconfiguring and Fine-Tuning
Chapter 4. Monitoring Arrays and Displaying System Statistics
Chapter 5. Identifying and Correcting Failures
Chapter 6. The Fibre Channel RAID Command-Line Interface
Chapter 7. The Fibre Channel Non-RAID Command-Line Interface
Appendix A. Error Recovery
Appendix B. Server Serial Connector Pinouts
Appendix C. Cache and Cache Page Size
Index
home/search |
what's new |
help
|
