A.3 Configuring the System Device to Use EVMS

This section discusses how to configure the system device during the Linux install to use EVMS as the volume manager. For the purposes of this install, a system device is any device that contains the Linux /boot, swap, or root (/) partitions for your Linux computer.

A.3.1 Before the Linux Install

This section discusses the following:

Storage Deployment Considerations for the System Device

Before you begin, please review the following:

The modified install procedures in this section describe how to install OES Linux with EVMS as the volume manager of the system device. The instructions assume the following:

  • You want to use EVMS on the system device.

  • You are configuring only the system device during the install.

  • Any other devices on the system are left as unallocated free space to be configured after the system is operating and performing as expected.

All other deployment scenarios can use the normal install with regard to the system device.

Storage Deployment Considerations for Data Devices

During the install, if you set up traditional Linux file systems on data devices where you also plan to use NSS volumes with EVMS, make sure to do the following:

  • Leave unallocated free space on the device to use for NSS volumes.

  • Configure the devices to use EVMS as the volume manager.

Data Loss Considerations for the System Device

During the install, when you modify the Partitioning settings to use EVMS instead of LVM, you must delete the recommended LVM partitions and partition tables and create new partitions and partition tables. This destroys all data on the disk.

IMPORTANT:To avoid data loss, it is best to use the modified install option only on a new device.

If you upgrade to OES Linux from an existing Linux server or from a NetWare server, remnants of the prior system and data volumes on the system device might prevent a smooth installation and can result in data loss, depending on what is stored on your system device.

  • When you repartition the system device during the install to use EVMS, the installation deletes all data on the device.

  • You might get unexpected warnings that EVMS-controlled system or boot partitions cannot be created because of remnants of the old system. It might be necessary to wipe (zero-out) the drive, then begin the installation again.

In either case, if you have data volumes stored on the existing system device, the data is destroyed.

If you have data volumes on the system device, take one or more of the following precautionary measures before you upgrade:

  • Before upgrading using the modified install, move the data volumes from the system device to another device.

  • If you cannot move the volumes, make a backup copy of the data volumes, so you can restore the data volumes later from a backup copy.

Hardware RAID Controller Considerations

If the server uses a hardware RAID controller with SCSI devices, and you plan to use space on it for NSS pools and volumes, make sure the RAID controller and SCSI devices use Write-Through cache management, not Write-Back cache management. Use the controller’s BIOS setup routine or configuration utility to enable Write-Through cache management when you set up the RAID.

IMPORTANT:Using Write-Back cache management on SCSI drives with a journaled file system such as NSS can actually degrade file system performance instead of improving it. For information, see Troubleshooting the NSS File System in the Novell Storage Services File System Administration Guide for OES.

Known Issue (OES Linux)

In the initial version of OES Linux, certain hardware configurations utilizing an HP Smart Array Controller, such as the 5i RAID Controller, rely on a naming scheme that is incompatible with the EVMS module found on the shipping OES Linux CD1 ISO image. This issue prevents using EVMS on the primary hard drive of these machines.

IMPORTANT:This issue is resolved in OES Linux SP1 and later.

For details on this issue and an updated OES CD1 ISO image that resolves this problem, see TID 2971770, OES install fails with EVMS Volume Manager.

After booting from the CD, you must continue installing the server from CD rather than using a network installation procedure, or the patched files are overwritten.

If you use the same size partitions as the previous partitions on the disk, the installation fails.

A.3.2 During the Linux Install

WARNING:The following procedure to install Linux with EVMS as the volume manager for the system device destroys all data on the system device.

To install Linux with EVMS as the volume manager for your boot and system partitions, you must modify the Partitioning configuration in the Installation Settings.

  1. Begin the SLES 9 install for OES.

    For information, see Installing Open Enterprise Server (OES) Linux.

  2. When the installation reaches the Installations Settings screen, delete the recommended partitions and the partition table on the system disk so that the device can be marked to use EVMS as the volume manager instead of LVM.

    1. In the list of Installation Settings, select Partitioning.

    2. In the Partitioning menu, select Create Custom Partition Setup, then click Next.

    3. Select Custom Partition - for Experts, then click Next to open the Expert Partitioner options.

    4. Select Expert > Delete Partition Table and Disk Label, then click Yes twice to continue through the Warning advisories.

      This deletes the recommended partitions and the partition table on the system disk.

  3. Create a primary partition on the system disk to use as the boot partition.

    1. Click Create.

    2. From the list of devices, select the device you want to use for the boot partition, such as /dev/hda, then click OK.

      If you have a single system disk, only one device is available and you are not prompted to choose the device.

    3. Select Primary Partition, then click OK.

    4. Select Format, then select the native Linux file system you want to use, such as Reiser.

    5. In Size (End Value) field, specify 200 MB or larger.

      For example, to set the size at 300 MB, type 300M.

    6. Set the mount point to /boot.

    7. Click OK.

      The partition appears as a logical device in the devices list, such as /dev/hda1.

  4. Create a second primary partition on the system disk to use for your swap and system volumes as follows:

    1. Click Create.

    2. From the list of devices, select the device you want to use for the second primary partition, such as /dev/hda, then click OK.

      If you have a single system disk, only one device is available and you are not prompted to choose the device.

    3. Select Primary Partition, then click OK.

    4. Select Do Not Format, then select Linux LVM (0x8E) from the list of file system IDs.

    5. In Size (End Value field), set the cylinder End value to 5 GB or larger, depending on the combined partition size you need to contain your system and swap volumes based on the following recommendations:

      IMPORTANT:Do not make the system partition larger than necessary. The remaining space on the system disk can be used to create NSS volumes or native Linux volumes that are managed by EVMS.

      • For your system volume, allow 2 GB (minimum) to 10 GB (recommended), depending on the OES services that you intend to install.

      • If you intend to create additional NSS volumes on the same physical disk, you must leave unpartitioned space available.

      • Set aside 128 MB or larger for the swap volume.

        Swap management is different for Linux kernel 2.4.10 and later. How much swap to add depends on the RAM size, the tasks that are planned for the system, and whether you want to make more virtual memory available than the RAM provides.

        Some swap (at least 128 MB) is good to have to minimize the risk of losing data when active processes run out of RAM space. Swap is not required for systems with more than 1 GB of RAM. You must have at least 1 GB of virtual memory (RAM plus swap) during the install, but if the swap is more than 2 GB, you might not be able install on some machines.

      • The total size should be the size you need for your system volume plus the size you need for your swap volume.

      For example, if you have a 20 GB hard drive with 2 GB of RAM and plan to install all of the OES services on the system volume, your system partition should be at least 11 GB. The remaining 9 GB should remain as free unpartitioned space that can be used for NSS volumes or other Linux partitions that you might want to create later.

    6. Click OK.

      The partition appears as a logical device in the devices list, such as /dev/hda2.

  5. Modify the volume management type from LVM to EVMS for the second primary partition you created in Step 4 as follows:

    1. At the bottom of the page, click EVMS.

      Available partitions for EVMS appear as devices under /dev/evms, such as /dev/evms/hda2.

    2. In the EVMS Configurator, select the LVM partition created in Step 4, then click Create Container.

    3. In the Create EVMS Container dialog box, select the partition, specify the container name (such as system), then click Add Volume to create the lvm/system container, where system is the container name.

    4. Click OK.

      The EVMS Configurator displays the lvm/system container you just created, its size, and free space.

  6. Create the swap volume in the lvm/system container as follows:

    1. Select lvm/system, then click Add.

    2. In the Create Logical Volume dialog box, select Format, then select Swap from the File System drop-down menu.

    3. Specify swap as the volume name.

    4. Specify 1 GB (recommended) for the swap volume.

      The swap size should be 128 MB or larger, with a recommended size of 1 GB. For an explanation of this recommendation, see Step 4.e.

    5. Specify the mount point as swap.

    6. Click OK.

  7. Create the system volume in the lvm/system container as follows:

    1. Select lvm/system, then click Add.

    2. In the Create Logical Volume dialog box, select Format, then select the file system to use from the File System drop-down menu, such as Reiser.

    3. In the Volume Name field, specify a volume name, such as sys_lx.

    4. Specify the Size of the system volume as the remaining space available in the lvm/system partition by clicking Max.

    5. Specify the mount point as / (root volume).

    6. Click OK.

  8. Click Next to return to the list of devices.

    Below is an example of the physical and logical devices that should be configured on your system. Your setup depends on the number of devices in the server and the sizes you choose for your partitions.

    Device

    Size

    F

    Type

    Mount

    Start

    End

    Used By

    /dev/hda

    149.0 GB

    6Y160p0

    0

    19456

    /dev/hda1

    305.9 MB

    F

    Linux Native (Reiser)

    /boot

    0

    38

    /dev/hda2

    20.0 GB

    Linux LVM

    39

    2649

    EVMS lvm/system

    /dev/hdb

    111.8 GB

    SP1203N

    0

    14595

    /dev/evms/lvm/system/sys_lx

    10.0 GB

    F

    EVMS

    /

    -

    -

    /dev/evms/lvm/system/swap

    1.0 GB

    F

    EVMS

    swap

    -

    -

  9. Click Next to return to the Installation Settings page.

    You can dismiss the message warning that you should not mix EVMS and non-EVMS partitions on the same device.

  10. From the Installations Settings screen, click Software > Details, then select NSS from the available software options.

  11. Continue with the OES installation.

    IMPORTANT:After the install is complete, make sure to perform the mandatory post-install configuration of the related system settings to ensure that the system device functions properly under EVMS. Otherwise, the system fails to boot properly.

    For information, see After the Linux Install.

A.3.3 After the Linux Install

After the OES installation is complete, you must perform the following tasks to ensure that the system device functions properly under EVMS:

Edit the /etc/fstab File

When you boot the system, the kernel reads the /etc/fstab file to identify which file systems should be mounted and then mounts them. This file contains a table of file system information about the root (/), /boot, and swap partitions plus other partitions and file systems you want to mount.

The /boot partition is separate from the EVMS container where you placed the root (/) and swap partitions and is not managed by EVMS at this time. However, in the following steps, you disable boot.lvm and boot.md, then enable boot.evms. In effect, this forces EVMS to scan all the partitions at boot time, including the /boot partition, and it activates /boot under the /dev/evms directory. Therefore, this makes /boot a partition that is discovered by EVMS at startup, and requires that the device be listed under /dev/evms in the fstab file so it can be found when booting with boot.evms.

After the install, you must edit the /etc/fstab file to modify the location of the /boot partition so it is under the /dev/evms directory. For example, change /dev/hda1 to /dev/evms/hda1. Replace hda1 with the device name you used for your /boot partition.

IMPORTANT:When working in the /etc/fstab file, do not leave any stray characters or spaces in the file. This is a configuration file, and it is highly sensitive to such mistakes.

  1. Open the /etc/fstab file in a text editor.

  2. Locate the line that contains the /boot partition.

    For example, if your /boot partition uses device hda1 and the Reiser file system, look for a line similar to this:

    /dev/hda1 /boot reiser defaults 1 1
    
  3. In the Device Name column, modify the location of the /boot partition from /dev to /dev/evms so it can be managed by EVMS. Modify only the device name by adding /evms to the path:

    /dev/evms/hda1 /boot reiser defaults 1 1

  4. Save the file.

    The changes do not take affect until the server is restarted. Do not reboot at this time.

  5. Continue with Disable boot.lvm and boot.md.

Disable boot.lvm and boot.md

Disable boot.lvm and boot.md so they do not run at boot time. EVMS now handles the boot.

  1. In YaST, click System > Runlevel Editor > Expert Mode.

  2. Select boot.lvm.

  3. Click Set/Reset > Disable the Service.

  4. Select boot.md.

  5. Click Set/Reset > Disable the Service.

  6. Click Finish, then click Yes.

    The changes do not take affect until the server is restarted. Do not reboot at this time.

  7. Continue with Enable the boot.evms Service.

Enable the boot.evms Service

The boot.evms service should be enabled automatically after the install, but you should verify that it is enabled.

  1. In YaST, click System > Runlevel Editor > Expert Mode.

  2. Select boot.evms.

  3. Click Set/Reset > Enable the Service.

    The B runlevel option is automatically selected.

  4. Click Finish, then click Yes.

    The changes do not take affect until the server is restarted. Do not reboot at this time.

  5. Continue with Edit the /etc/init.d/boot.evms Script.

Edit the /etc/init.d/boot.evms Script

  1. Open the /etc/init.d/boot.evms script in a text editor.

  2. Add the following lines to the Stop section:

    mount -n -o remount,rw /
    
    echo -en "\nDeleting devices nodes"
    
    rm -rf /dev/evms
    
    mount -n -o remount,ro /
    

    For example, the Stop section looks like this after the edit:

    stop)
    
        echo -n "Stopping EVMS"
    
        mount -n -o remount,rw /
    
        echo -en "\nDeleting devices nodes"
    
        rm -rf /dev/evms
    
        mount -n -o remount,ro /
    
        rc_status -v
    
      ;;
    
  3. Save the file.

  4. Continue with Enable Write-Through Cache Management for SCSI Devices.

Enable Write-Through Cache Management for SCSI Devices

If you are using space from SCSI devices for your NSS pools and volumes on Linux, make sure the devices use Write-Through cache management, not Write-Back cache management to avoid possible data loss in the event of power loss. Write-Through and Write-Back (also known as Write-Behind) are cache management schemes that can affect the manner in which data is written to the hard drives. Although Write-Back can sometimes improve performance, it does so by allowing data to be held in cache rather than being written to disk.

IMPORTANT:If your device’s controller board provides an on-board backup power supply, it could potentially prevent some data loss. It is up to you to determine if the power backup is sufficient for power loss scenarios in your production environment and whether you want to take the risk of losing data if it is not.

If the server uses a RAID controller, use the controller’s BIOS setup routine or configuration utility to configure Write-Through cache management.

To enable Write-Through cache management for local devices:

  1. Log in to the server as root.

  2. If the scsi-config utility is not already installed. install it using the xscsi RPM.

    The xscsi RPM, which contains the scsi-config command, is not installed by default.

    1. In YaST, open the Various Linux Tools section.

    2. Install the xscsi RPM package, then close YaST.

      The xscsi package installs the scsi-config utility in /user/bin/scsi-config.

  3. Enable Write-Through (disable Write-Back) cache management for each SCSI device by performing the following for each device where you plan to use NSS volumes.

    1. In a terminal console, enter

      scsi-config 
      
    2. In the window that opens, browse to select drive you want to manage, then click Continue.

    3. Click Cache Control Page.

    4. Enable Write-Through cache management mode by deselecting the Write cache enabled check box.

      Write-Through cache management is enabled by default, so the Write cache enabled check box should be deselected. If the Write cache enabled check box is selected, Write-Back cache management mode is enabled and you deselect the box to disable Write-Back cache management.

    5. Click Quit > Save Changes.

    6. When prompted to confirm the change, click Go Ahead and Save > Quit.

    7. To verify the setting, at a terminal console prompt, enter

      scsiinfo -c /dev/sdx
      

      Replace /dev/sdx with the device you are checking.

      A value of 0 for Write Cache means that the drive is in Write-Through cache management mode.

SCSI Devices in a Hardware RAID

If you have not already done so, use the RAID controller’s BIOS setup routine or configuration utility to enable Write-Through cache management for the controller and the RAID’s SCSI devices.

Local SCSI Devices

To enable Write-Through cache management for local devices:

  1. Log in to the server as root.

  2. In YaST, open Software > Install and Remove Software > Various Linux Tools.

  3. Install the xscsi RPM package, then close YaST.

    The xscsi package installs the scsi-config utility in /user/bin/scsi-config.

  4. In a terminal console, enter scsi-config, then enable Write-Through (disable Write-Back) for each SCSI device.

Reboot the Server

  1. Reboot the server to apply the post-install configuration settings.

Verify the System Services

After the post-install configuration is complete and you have rebooted the server, make sure the server is operating as expected. For example:

  • Check the server health. For information, see Monitoring Server Health in the Novell Remote Manager Administration Guide for OES Linux.

  • Make sure the free space on the system device is available to you in iManager. For information about how to view devices in iManager, see Viewing Devices on a Server in the Novell Storage Services File System Administration Guide for OES.

  • Make sure that NSS modules are running. In YaST, click System > Runlevel Editor > Expert Mode. The NSS module should be enabled. Any changes you make do not take effect until you reboot the server.

Create NSS Pools and Volumes

For any data devices where you plan to create NSS volumes, make sure to create at least one NSS volume on the device before you create any traditional Linux volumes. You must use the Storage management plug-in to iManager or NSSMU to create the pools and volumes. These tools automatically configure the devices to use EVMS and configure the partitions for the NSS pools and volumes that you create. The order of creation and the toolset you use helps ensure that EVMS automatically controls the device.

For information, see the Novell Storage Services File System Administration Guide for OES.