General Parallel File System 3.

2 Labs
July 7, 2010

CONTENTS
Lab 1 1a 1b 2 3 4 5 6 Title Install and configure a GPFS cluster Install and configure a GPFS cluster on AIX via Command Line Guide to Installing GPFS* Storage Pools, Filesets and Policies Using Replication Snapshots Dynamically Adding a Disk to an Online File System Permanently Remove GPFS* Pg 3 7 14 21 28 31 33 35

*This was not produced by IBM. Some labs were modified to promote consistency, or correct errors. Other labs were adapted from other IBM documentation.

GPFS Labs Exercise 1 Install and configure a GPFS cluster

Objectives: Use the GPFS web based administration tool to install a GPFS cluster. Requirements: Node names and IP address provided by instructor Node1:________________________ Node2:________________________ Account/Cluster name used for exercise Name: root Password: __________________ ClusterName: _______________ Step 1: Initialize the Lab Environment This lab assumes the rpm packages have already been installed on the servers. The packages for linux are:
gpfs.base-3.2.1-0.i386.rpm gpfs.docs-3.2.1-0.noarch.rpm gpfs.gui-3.2.1-0.i386.rpm gpfs.src-3.2.1-0.noarch.rpm gpfs.base-3.2.1-0.i386.update.rpm gpfs.gpl-3.2.1-0.noarch.rpm gpfs.msg.en_US-3.2.1-0.noarch.rpm

You can install the packages manually by copying them to each node and running rpm or the GPFS install wizard will distribute the packages for you. You need to manually install the packages on at least one node and start the GUI by running:
/etc/init.d/gpfsgui start

Step 2: Create the GPFS cluster In this step you will create a GPFS cluster on two nodes using the GPFS web based administration interface. 1 Open a web browser to the GPFS http://\[node1 ip address\]/ibm/console 2 Login using the account and password provided by the instructor.

Account: root Password: (root password) 3 In the navigation pane select "GPFS Management" then "Install GPFS". This will start the GPFS configuration wizard.

4 5 a.

Select "Create a new Session". This will take you to the Define hosts page. Under "Defined hosts" click "Add" Enter the name or ip address of node1, and the root password.

b. Select "Add this host and add the next host."

6.

Close the Task process dialogue when completed.

7 8

Enter the IP address and root password for node2 and select "Add host." Close the Task process dialogue when completed. The hosts have now been added. Select next to go to the Install and Verify Packages page. On this page select "Check existing package installation." Close the task dialog when the check is complete. GPFS Ships an open source component called the GPL layer that allows the support of a wide variety of Linux kernels. The GPL layer installation page checks that the GPL layer is built and installed correctly. If it is not, the installer will complete the build and install. Select "Check existing GPL layer installation". Close the "Task" dialog when the check is complete. GPFS verifies the network configuration of all nodes in the cluster. Select "Check current settings" to verify the network configuration. Close the "Task" dialog when the check is complete. GPFS uses ssh (Or other remote command tool) for some cluster operations. The installer will verify that ssh is configured properly for a GPFS cluster. Select "Check Current Settings" to verify the ssh config. Close the "Task" dialog when the check is complete. It is recommended, though not required, that all the servers synchronize the time using a protocol such as NTP. For this lab we will skip the NTP setup. Choose "Skip Setup" to continue.

10

11

12

13 14 15 16

Next, you set the name the GPFS cluster. Enter the cluster name and select "Next." The last step is to define the primary and secondary cluster configuration servers. Since this is a two node cluster we will leave it at the defaults. Select "Next" to continue. Select Next to complete the cluster configuration. Close the "Task" dialog when the configuration is complete. When you select "Finish" you will be directed to the cluster management page.

17

The GPFS cluster is now installed and running.

GPFS Labs Exercise 1a Install and configure a GPFS cluster on AIX via Command Line
Objectives Verify the system environment Create a GPFS cluster Define NSD's Create a GPFS file system

Requirements An AIX 5.3 System o Very similar to linux config with rpm instead of AIX binary images and Linux admin commands are different At least 4 hdisks GPFS 3.3 Software with latest PTF

Step 1: Verify Environment 1 Verify nodes properly installed a Check that the oslevel is supported On the system run oslevel Check the GPFS FAQ:http://publib.boulder.ibm.com/infocenter/clresctr/vxrx/index.jsp__ b Is the installed OS level supported by GPFS? Yes No c Is there a specific GPFS patch level required for the installed OS? Yes No d If so what patch level is required? ___________ 2 Verify nodes configured properly on the network(s) a Write the name of Node1: ____________ b Write the name of Node2: ____________ c From node 1 ping node 2 d From node 2 ping node 1 If the pings fail, resolve the issue before continuing. 3 Verify node-to-node ssh communications (For this lab you will use ssh and scp for communications) 7

a On each node create an ssh-key. To do this use the command ssh-keygen; if you don't specify a blank passphrase, -N, then you need to press enter each time you are promoted to create a key with no passphrase until you are returned to a prompt. The result should look something like this:
# ssh-keygen -t rsa -N "" -f $HOME/.ssh/id_rsa Generating public/private rsa key pair. Created directory '/.ssh'. Your identification has been saved in /.ssh/id_rsa. Your public key has been saved in /.ssh/id_rsa.pub. The key fingerprint is: 7d:06:95:45:9d:7b:7a:6c:64:48:70:2d:cb:78:ed:61 sas@perf3-c2-aix

b On node1 copy the /.ssh/id_rsa.pub file to /.ssh/authorized_keys

# cp /.ssh/id_rsa.pub /.ssh/authorized_keys

c From node1 copy the /.ssh/id_rsa.pub file from node2 to /tmp/id_rsa.pub

# scp node2:/.ssh/id_rsa.pub /tmp/id_rsa.pub

d Add the public key from node2 to the authorized_keys file on node1
# cat /tmp/id_rsa.pub >> /.ssh/authorized_keys

e Copy the authorized key file from node1 to node2

# scp /.ssh/authorized_keys node2:/.ssh/authorized_keys

To test your ssh configuration ssh as root from node 1 to node1 and node1 to node2 until you are no longer prompted for a password or for addition to the known_hosts file.
node1# node1# node2# node2# ssh ssh ssh ssh node1 node2 node1 node2 date date date date

g Supress ssh banners by creating a .hushlogin file in the root home directory
# touch /.hushlogin

Verify the disks are available to the system For this lab you should have 4 disks available for use hdiskn-hdiskt. 1. Use lspv to verify the disks exist 2. Ensure you see 4 disks besides hdisk0 talk.

Step 2: Install the GPFS software On node1: 1 Locate the GPFS software in /yourdir/software/base/
# cd /yourdir/software/base/

Run the inutoc command to create the table of contents

# inutoc .

Install the base GPFS code using the installp command

# installp -aXY -d/yourDir/software/basegpfs -f all

Locate the latest GPFS patch level in /yourdir/software/PTF/

# cd /yourdir/software/PTF/

Run the inutoc command to create the table of contents

# inutoc .

Install the PTF GPFS code using the installp command

# installp -aXY -d/yourdir/software/PTFgpfs -f all

Repeat Steps 1-7 on node28. On node1 and node2 confirm GPFS is installed using lslpp
# lslpp -L gpfs.\*

the output should look similar to this

# lslpp -L gpfs.\* Fileset Level State Type Description (Uninstaller) ---------------------------------------------------------------------------gpfs.base 3.3.0.3 A F GPFS File Manager gpfs.docs.data 3.3.0.3 A F GPFS Server Manpages and Documentation gpfs.gui 3.3.0.3 C F GPFS GUI gpfs.msg.en_US 3.3.0.1 A F GPFS Server Messages U.S. English

Note: Exact versions of GPFS may vary from this example, the important part is that all three packages are present.

Confirm the GPFS binaries are in your path using the mmlscluster command
# mmlscluster mmlscluster: 6027-1382 This node does not belong to a GPFS cluster. mmlscluster: 6027-1639 Command failed. Examine previous error messages to determine cause.

Note: The path to the GPFS binaries is: /usr/lpp/mmfs/bin Step 3: Create the GPFS cluster For this exercise the cluster is initially created with a single node. When creating the cluster make node1 the primary configuration server and give node1 the designations quorum and manager. Use ssh and scp as the remote shell and remote file copy commands. Primary Configuration server (node1): ________ Verify fully qualified path to ssh and scp: ssh path________ scp path_____________ 1 Use the mmcrcluster command to create the cluster
# mmcrcluster -N _node01_:manager-quorum -p _node01_ -r /usr/bin/ssh -R /usr/bin/scp

Run the mmlscluster command again to see that the cluster was created
# mmlscluster GPFS cluster information ======================== GPFS cluster name: GPFS cluster id: GPFS UID domain: Remote shell command: Remote file copy command: node1.ibm.com 13882390374179224464 node1.ibm.com /usr/bin/ssh /usr/bin/scp

GPFS cluster configuration servers: ----------------------------------Primary server: Secondary server: node1.ibm.com (none)

Node Daemon node name IP address Admin node name Designation ----------------------------------------------------------------------------------------------1 perf3-c2-aix.bvnssg.net 10.0.0.1 node1.ibm.com quorum-manager

Set the license mode for the node using the mmchlicense command. Use a server license for this node.
# mmchlicense server --accept -N node01

Step 4: Start GPFS and verify the status of all nodes 1 Start GPFS on all the nodes in the GPFS cluster using the mmstartup command
# mmstartup a

10

Check the status of the cluster using the mmgetstate command

# mmgetstate -a Node number Node name GPFS state -----------------------------------------1 node1 active

Step 5: Add the second node to the cluster 1 One node 1 use the mmaddnode command to add node2 to the cluster
# mmaddnode -N node2

Confirm the node was added to the cluster using the mmlscluster command
# mmlscluster

Use the mmchcluster command to set node2 as the secondary configuration server
# mmchcluster -s node2

Set the license mode for the node using the mmchlicense command. Use a server license for this node.
# mmchlicense server --accept -N node02

Start node2 using the mmstartup command

# mmstartup -N node2

Use the mmgetstate command to verify that both nodes are in the active state
# mmgetstate -a

Step 6: Collect information about the cluster Now we will take a moment to check a few things about the cluster. Examine the cluster configuration using the mmlscluster command 1. What is the cluster name? ______________________ 2. What is the IP address of node2? _____________________ 3. What date was this version of GPFS "Built"? ________________ Hint: look in the GPFS log file: /var/adm/ras/mmfs.log.latest Step 7: Create NSDs 11

You will use the 4 hdisks. 1 2 Make sure they can all hold data and metadata Leave the storage pool column blank. Leave the Primary and Backup server fields blank Sample input files are in /yourdir/samples

On node 1 create directory /yourdir/data Create a disk descriptor file /yourdir/data/diskdesc.txt using the format:
#DiskName:PrimaryServer:BackupServer:DiskUsage:FailureGroup:DesiredName:StoragePool hdiskw:::dataAndMetadata::nsd1: hdiskx:::dataAndMetadata::nsd2: hdisky:::dataAndMetadata::nsd3: hdiskz:::dataAndMetadata::nsd4:

Note: hdisk numbers will vary per system. 3 Create a backup copy of the disk descriptor file /yourdir/data/diskdesc_bak.txt
# cp /yourdir/data/diskdesc.txt /yourdir/data/diskdesc_bak.txt

Create the NSD's using the mmcrnsd command

# mmcrnsd -F /yourdir/data/diskdesc.txt

Step 8: Collect information about the NSD's Now collect some information about the NSD's you have created. 1. Examine the NSD configuration using the mmlsnsd command 1. What mmlsnsd flag do you use to see the operating system device (/dev/hdisk?) associated with an NSD? _______ Step 9: Create a file system Now that there is a GPFS cluster and some NSD's available you can create a file system. In this section we will create a file system. Set the file system blocksize to 64kb Mount the file system at /gpfs

12

Create the file system using the mmcrfs command

# mmcrfs /gpfs fs1 -F diskdesc.txt -B 64k

Verify the file system was created correctly using the mmlsfs command
# mmlsfs fs1

Is the file system automatically mounted when GPFS starts? _______________ 3 Mount the file system using the _mmmount_ command
# mmmount all a

Verify the file system is mounted using the df command

# df -k Filesystem /dev/hd4 /dev/hd2 /dev/hd9var /dev/hd3 /dev/hd1 /proc /dev/hd10opt /dev/fs1 1024-blocks Free %Used 65536 6508 91% 1769472 465416 74% 131072 75660 43% 196608 192864 2% 65536 65144 1% 327680 47572 86% 398929107 398929000 1% Iused %Iused Mounted on 3375 64% / 35508 24% /usr 620 4% /var 37 1% /tmp 13 1% /home - /proc 7766 41% /opt 1 1% /gpfs

Use the mmdf command to get information on the file system.

# mmdf fs1

How many inodes are currently used in the file system? ______________

13

GPFS Labs Exercise 1b Guide to Installing GPFS on Red Hat/CentOS 5.x

This document describes installing GPFS 3.2 on Red Hat/CentOS 5 systems. There are two sets of files to install from. The first is the original installer. It is a self-extracting archive containing the RPMs for version 3.2.0-0. The second set contains the four RPMs from that same archive. You may choose to use either method. When you are finished with this guide, you will have a GPFS cluster up and running. You will not have storage attached to the cluster, yet. That will be the next step. Following is an explanation of the files: gpfs.base-3.2.0-0.i386.rpm This is the main GPFS code, including daemons and utility binaries. gpfs.docs-3.2.0-0.noarch.rpm Obviously documentation. Installing this is optional. gpfs.gpl-3.2.0-0.noarch.rpm Portability modules for Linux systems. gpfs.msg.en_US-3.2.0-0.noarch.rpm This is the US English language files. Optional, but you will want to install this. There are also five updates. These include an update file at version 3.2.1-20 for each of files listed above, as well as an update file for the GUI (web management interface). Update files cannot be installed without first installing the 3.2.0-0 versions, with the exception of the gpfs.gui file; it may be installed without a 3.2.0-0 version of the same. Objectives Install GPFS with updates to a Red Hat/CentOS 5 server Create a new GPFS cluster with a primary and secondary manager and one quorum node Add additional nodes into the GPFS cluster

Requirements GPFS 3.2.0-0 software GPFS 3.2.1-20 updates (optional) Red Hat/CentOS 5 servers in the same subnet with Internet access (for performing yum updates)

Step 1: Installing the RPMs Installing from the IBM Installer If you are using the self-extracting installer, perform the following to extract the RPMs:

14

# gpfs_install-3.2.0-0_arch

where arch is your system architecture (i386 or x64). The following 3.2.0-0 RPMs will end up in the /usr/lpp/mmfs/3.2 directory: gpfs.base-3.2.0-0.arch.rpm gpfs.docs-3.2.0-0.noarch.rpm gpfs.gpl-3.2.0-0.noarch.rpm gpfs.msg.en_US-3.2.0-0.noarch.rpm Note that you will need to have XWindows running, as well as IBMs JRE (Sun and Microsoft JREs are insufficient) to view and accept the license agreement. The --text-only switch is reported by IBM to remove the XWinows requirement, however the JRE is still required. Using Extracted Files You may have files that were previously extracted and stored on a medium. If you are using these extracted files, simply copy the files to a directory from which you can install the RPMs. Now, copy the updated 3.2.1-20 RPMs into a work directory from which you can install them. Step 2: Installing GPFS This covers installing the files, which is not the same as installing the cluster. The files get everything in place, whereas installing the cluster (covered below) implies configuring and starting GPFS. Begin by installing the compat-libstdc++-33 package. This is a dependency for the gpfs.base package:
# yum install compat-libstdc++-33

To install the 3.2.0-0 RPMs, perform the following:

# # # # rpm rpm rpm rpm ivh ivh ivh ivh gpfs.base-3.2.0-0.i386.rpm gpfs.docs-3.2.0-0.noarch.rpm gpfs.gpl-3.2.0-0.noarch.rpm gpfs.msg.en_US-3.2.0-0.noarch.rpm

To install the 3.2.1-21 RPMs, perform the following:

# # # # # rpm rpm rpm rpm rpm Uvh Uvh Uvh Uvh Uvh gpfs.base-3.2.1-20.i386.update.rpm gpfs.docs-3.2.1-20.noarch.rpm gpfs.gpl-3.2.1-20.noarch.rpm gpfs.msg.en_US-3.2.1-20.noarch.rpm gpfs.gui-3.2.1-20.noarch.rpm

Finally, start the gui application framework:

15

# /etc/init.d/gpfsgui start

This will install GPFS to /usr/lpp/mmfs. The steps above must be completed for every PC in the cluster with the exception of the gpfs.gui commands. These may be performed on only one host in the cluster that can act as a a manager node (presumably, the first node which is identified as the primary below). Note that the gpfsgui is also a service, and will start automatically the next time the server starts. Step 3: Compiling and Installing the Portability Modules Linux installations require that some modules be compiled and installed. IBM refers to these as portability modules. The module source code is installed in /usr/lpp/mmfs/src. Alternately, the updated modules may be downloaded from the DeveloperWorks site. Before compiling, you will require the applications necessary to compile software against your current kernel. For Red Hat/CentOS, the following will suffice:
# yum install kernel kernel-headers kernel-devel gcc gcc-c++ imake

Once complete, it's good practice to verify that the /boot/grub/grub.conf file is referencing the new versions of the kernel and initrd files as the default. Reboot the host so that the new kernel can take effect. For CentOS (or other distributions that are based on Red Hat, but are not Red Hat strictly), you will need to edit the /etc/redhat-release file. This file contains the name and version of the operating system. When configuring, the configure script looks for Red Hat Enterprise Linux in this file, and will fail if it is not found. Using your favorite editor, change:
CentOS release 5.x

to

Red Hat Enterprise Linux release 5.x

Next, you will need to set the SHARKCLONEROOT environment variable. The IBM documentation indicates that if not set, it will default to /usr/lpp/mmfs/src, however the configure script will terminate with an error if the environment variable is not set.
# export SHARKCLONEROOT=/usr/lpp/mmfs/src

Configure the build with:

# cd /usr/lpp/mmfs/src/config # ./configure

The configure script should end with no output to the terminal. Finally, you can begin building the modules. Return to the parent directory (/usr/lpp/mmfs/src) and issue the following to compile and install the modules:

16

# cd .. # make World # make InstallImages

Step 4: Creating the Cluster and Adding Nodes There are two methods to creating a GPFS cluster, through the GUI, and at the command line. Little success was experienced using the GUI with the versions installed above. This document covers strictly the command line. That said, the GUI does come in useful for some troubleshooting activities. If needed, connect to the GUI using your web browser going to the following address: http://node1_ipaddress/ibm/console/ The username and password will likely be the root user and password, unless you have made special provisions for another account with superuser access during the installation. The instructions below use nodeX-fqdn to refer to hosts in the cluster. This assumes that DNS is working. If you are in an environment without DNS, you have the option of either modifying host files on each host to achieve the same goal, or using IP addresses instead of hostnames. Regardless of the method you choose, that method should be reflected when setting up ssh keys so that the authorized host in the key matches that used to refer to hosts. Enable Passwordless Login to Each GPFS Host from Each GPFS Host Using ssh Keys First create a key on each host with the following:
# ssh-keygen t rsa N "" f ~/.ssh/id_rsa Generating public/private rsa key pair. Created directory /.ssh Your identification has been saved in /home/username/.ssh/id_rsa. The key fingerprint is: 01:23:45:67:89:ab:cd:ef:01:23:45:67:89:ab:cd:ef user@host.example.com

Note: If you prefer to use DSA over RSA, simply use the following:
# ssh-keygen t dsa N "" f ~/.ssh/id_dsa

The same information is echoed back to the screen, and you will end up with id_dsa instead of id_rsa in the .ssh directory. Place the directory containing GPFS commands in roots path by adding:
;/usr/lpp/mmfs/bin

17

To the end of the PATH= line in ~/bash_profile. (The ; is required to separate this path specification from the last one In the line.) Log out, then back in for the changes to take effect (or simply make the same change in a shell window). Now, transfer the public key to every other host in the cluster:
# cat ~/.ssh/id_rsa.pub | ssh user@server "cat - >> \ ~/.ssh/authorized_keys"

Replace user with root and replace server with the fqdn of the remote host, and issue the command a second time replacing server with the IP address of the remote host. Perform these steps on every host in the cluster, so root on each host can log into the root account on every other host without the requirement of a password. In addition, you must remove any banner text that may be displayed through motd, or the ssh Banner parameter during login. Start and Add a Node to the Cluster Verify that GPFS is properly installed by performing the following:
# mmlscluster

This is produce an error message, but will insure that the cluster software is installed and available. Verify the location of the ssh and scp binaries. This will be used below.
# which ssh # which scp

Create the first cluster node:

# mmcrcluster N node1-fqdn:manager-quorum p node1-fqdn r /usr/bin/ssh \ R /usr/bin/scp

1. The N node1-fqdn indicates that you are creating a node on the host identified by node1-fqdn (presumably, the current host). 2. The manager-quorum indicates that this is a manager node (it retains copies of the configuration files) and a quorum node. 3. The p node1-fqdn indicates that this node will be the Primary manager. 4. The remainder specifies which binaries will be used for remote command execution and file transfers. Run the mmlscluster command again:
# mmlscluster

18

This should respond with a screen of information about the cluster, including some of the detail provided above. Prior to installing the remaining nodes, you need to start the cluster:
# mmstartup a # mmgetstate a Node number Node name GPFS state ----------------------------------------1 node1 active

The first command starts the cluster. The second displays a status of all (through a) nodes in the cluster (of which there are currently only one). Add Remaining Nodes to the Cluster
# # # # # mmaddnode N node2-fqdn mmlscluster mmchcluster s node2-fqdn mmstartup N node2-fqdn mmgetstate a

The commands above do the following: 1. Add a second node to the cluster 2. List all nodes in the cluster 3. Set node2-fqdn as the Secondary manager node 4. Start GPFS on the second node 5. Display the status of all nodes in the cluster For additional nodes, follow the same instructions, omitting the mmchcluster s command. At this point, the cluster should be up and running, and all of the GPFS nodes added. You can now begin adding storage to the cluster. Troubleshooting Verify Passwordless Login GPFS exchanges files and executes commands on all servers in the cluster. This requires that each server have access to other servers without the need to type a password. If you cannot log into another host with the root account, and without a password, GPFS will not be able to communicate properly with other nodes. For example, if you are on node1, and you want to test ssh to node2, perform the following:
node1# ssh node2 node2#

19

Notice that there was no prompt for a password, and no motd/banner messages appear on the terminal before a prompt is issued. Use the GUI as a Troubleshooting Tool You can also use the GUI as a troubleshooting tool. If you have a host, node3-fqdn that is having trouble, Launch the GUI. Select GPFS Management|Install GPFS|Add host. Type node2-fqdn in the host name field, and the root password for node2-fqdn. Click the Add Host button. A window will appear with three stages of action to be performed. Connectivity Check fails: This indicates that the current host cannot communicate with node3fqdn. That may be that the password is incorrect, sshd is not running on the remote host, or passwordless authentication (using ssh keys) is not set up correctly. Gathering system information fails: This indicates that the portability modules (kernel modules) either were not installed, or were not installed correctly on node3-fqdn. Cluster membership check fails: This indicates the node is already a member of another cluster. Nodes in an unknown State I shut all of my GPFS hosts down, and when I restart, only the Primary manager is active (when running the mmgetstate a command). All of the others are reported as unknown. Problem: Unknown. This has been observed in CentOS5 + VMWare + Open/iSCSI environments. Solution: You will need to drop all of the unknown nodes out of the cluster then re-add them. This will have the effect of renumbering the nodes, however. This solution is a particular problem for node2 which (from the instructions above) was installed as the secondary manager. You cannot remove a primary or secondary manager from the cluster. The strategy is to shuffle around which node is the secondary manager during the drop and add process. The following will accomplish this for you. In this example, there are four nodes, node1 which is the primary manager (and is active), node2 which is the secondary manager, and node3/node4 which have no special designation.
# # # # # # # # # # mmshutdown a mmdeletenode N node3 mmdeletenode N node4 mmaddnode N node3 mmaddnode N node4 mmchcluster s node3 (changes the sec. manager to node 3) mmdeletenode N node2 mmaddnode N node2 mmchcluster s node2 mmstartup a

20

GPFS Lab Exercise 2 Storage Pools, Filesets and Policies

Objectives: * Create storage pools * Create a file system * Create filesets * Implement a Placement Policy * Define and execute a file management policy Requirements: Complete Exercise 1: Installing the cluster List of available devices /dev/sd__ /dev/sd__ /dev/sd__ /dev/sd__ /dev/sd__ Step 1: Create A File System With 2 Storage Pools Storage pools are defined when an NSD (Network Shared Disk) is created. You will use the 4 disks (labelled sd_ through sd_ ) provided by your instructor to create 2 storage pools. Since the storage is all direct attached, no NSD servers are needed. Place the disks into two storage pools and make sure both pools can store file data. Hints: Create 2 disks per storage pool The first pool must be the system pool. The system pool should be able to store data and metadata Only the system pool can contain metadata # Create a disk descriptor file /gpfscourse/data/pooldesc.txt using the format:#DiskName:serverlist::DiskUsage:FailureGroup:DesiredName:StoragePool
/dev/sd__:::dataAndMetadata::nsd1:system /dev/sd__:::dataAndMetadata::nsd2:system /dev/sd__:::dataOnly::nsd3:pool1 /dev/sd__:::dataOnly::nsd4:pool1

The two storage pools will be system and pool1. 1 Create a backup copy of the disk descriptor file 21

/gpfs-course/data/pooldesc_bak.txt

Create the NSD's using the mmcrnsd command.

> mmcrnsd -F /gpfs-course/data/pooldesc.txt -v no mmcrnsd: Processing disk sdc mmcrnsd: Processing disk sdd mmcrnsd: Processing disk sde mmcrnsd: Processing disk sdf mmcrnsd: Propagating the cluster configuration data to all affected nodes. This is an asynchronous process.

Create a file system based on these NSD's using the mmcrfs command * Set the file system blocksize to 64KB * Mount the file system at /gpfs Command: "mmcrfs /gpfs fs1 -F /gpfs-course/data/pooldesc.txt -B 64k -M 2 -R 2" Example:
gpfs1:~ # mmcrfs /gpfs fs1 -F /gpfs-course/data/pooldesc.txt -B 64k -M2 -R2 The following disks of fs1 will be formatted on node gpfs1: nsd1: size 20971520 KB nsd2: size 20971520 KB nsd3: size 20971520 KB nsd4: size 20971520 KB Formatting file system ... Disks up to size 53 GB can be added to storage pool 'system'. Disks up to size 53 GB can be added to storage pool 'pool1'. Creating Inode File 45 % complete on Wed Sep 26 10:05:27 2007 89 % complete on Wed Sep 26 10:05:32 2007 100 % complete on Wed Sep 26 10:05:33 2007 Creating Allocation Maps Clearing Inode Allocation Map Clearing Block Allocation Map 42 % complete on Wed Sep 26 10:05:52 2007 83 % complete on Wed Sep 26 10:05:57 2007 100 % complete on Wed Sep 26 10:05:59 2007 43 % complete on Wed Sep 26 10:06:04 2007 85 % complete on Wed Sep 26 10:06:09 2007 100 % complete on Wed Sep 26 10:06:10 2007 Completed creation of file system /dev/fs1. mmcrfs: Propagating the cluster configuration data to all affected nodes. This is an asynchronous process.

Verify the file system was created correctly using the mmlsfs command
# mmlsfs fs1

Mount the file system using the mmmount command

# mmmount fs1 -a

Verify the file system is mounted using the df command.

# df

22

Filesystem /dev/sda3 tmpfs /dev/sda1 /dev/fs1

1K-blocks Used Available Use% Mounted on 32908108 10975916 21932192 34% / 2019884 4 2019880 1% /dev/shm 72248 45048 27200 63% /boot 3989291072 491992640 3497298432 13% /gpfs

Verify the storage pool configuration using the mmdf command

# mmdf fs1 disk disk size failure holds holds free KB free KB name in KB group metadata data in full blocks in fragments --------------- ------------- -------- -------- ----- -------------------- ------------------Disks in storage pool: system nsd1 102734400 -1 yes yes 102565184 (100%) 90 ( 0%) nsd2 102734400 -1 yes yes 102564608 (100%) 96 ( 0%) -------------------------------- ------------------(pool total) 205468800 205129792 (100%) 186 ( 0%) Disks in storage pool: pool1 nsd3 102734400 nsd4 102734400 ------------(pool total) 205468800 (data) (metadata) (total) ============= 410937600 205468800 ============= 410937600 -1 no -1 no yes yes 102732288 (100%) 62 ( 0%) 102732288 (100%) 62 ( 0%) -------------------- ------------------205464576 (100%) 124 ( 0%) ==================== =================== 410594368 (100%) 310 ( 0%) 205129792 (100%) 186 ( 0%) ==================== =================== 410594368 (100%) 310 ( 0%)

Inode Information ----------------Number of used inodes: Number of free inodes: Number of allocated inodes: Maximum number of inodes:

4038 397370 401408 401408

Step 2: Create Filesets We are going to create 5 filesets to organize the data. 1 Create 5 filesets fileset1-fileset5 using the mmcrfileset command
# # # # # mmcrfileset mmcrfileset mmcrfileset mmcrfileset mmcrfileset fs1 fs1 fs1 fs1 fs1 fileset1 fileset2 fileset3 fileset4 fileset5

Verify they were created using the mmlsfileset command

# mmlsfileset fs1

What is the status of fileset1-fileset5? _______________ 3 Link the filesets into the file system using the mmlinkfileset command
# # # # # mmlinkfileset mmlinkfileset mmlinkfileset mmlinkfileset mmlinkfileset fs1 fs1 fs1 fs1 fs1 fileset1 fileset2 fileset3 fileset4 fileset5 -J -J -J -J -J /gpfs/fileset1 /gpfs/fileset2 /gpfs/fileset3 /gpfs/fileset4 /gpfs/fileset5

23

Now what is the status of fileset1-fileset5? ___________________ Step 3: Create a file placement policy Now that you have two storage pools and some filesets you need to define placement policies to instruct GPFS where you would like the file data placed. By default, if the system pool can accept data, all files will go to the system storage pool. You are going to change the default and create 3 placement rules. The rules will designate:
1

Data in fileset1 to fileset4 go to the system storage pool Data in fileset5 go to the pool1 storage pool files that end in .dat go to pool1 storage pool Start by creating a policy file /gpfs-course/data/placementpolicy.txt
/* The fileset does not matter, we want all .dat and .DAT files to go to pool1 */ RULE 'datfiles' SET POOL 'pool1' WHERE UPPER(name) like '%.DAT' /* All non *.dat files placed in filset5 will go to pool1 */ RULE 'fs5' SET POOL 'pool1' FOR FILESET ('fileset5') /* Set a default rule that sends all files not meeting the other criteria to the system pool */ RULE 'default' set POOL 'system'

Install the new policy file using the mmchpolicy command

# mmchpolicy fs1 placementpolicy.txt

Step 4: Testing the placement policies Now you will do some experiments to see how policies work. Use this chart to track the experiments results. You can get the amount of free space by using the mmdf command. Experiment Before Bigilfe1 bigfile1.dat bigfile2 Migrate/Delete 1 2 Record the "Before" free space in the chart Create a file in fileset1 called bigfile1
# dd if=/dev/zero of=/gpfs/fileset1/bigfile1 bs=64k count=10000

System Pool (Free KB)

Pool1 (Free KB)

Record the free space in each pool using the mmdf command (Bigfile1) 24

# mmdf fs1 disk disk size failure holds name in KB group metadata --------------- ------------- -------- -------Disks in storage pool: system nsd1 20971520 -1 yes nsd2 20971520 -1 yes ------------(pool total) 41943040 Disks in storage pool: pool1 nsd3 20971520 nsd4 20971520 ------------(pool total) 41943040 (data) (metadata) (total) ============= 83886080 41943040 ============= 83886080 -1 no -1 no

holds free KB free KB data in full blocks in fragments ----- -------------------- ------------------yes yes 20588288 ( 98%) 930 ( 0%) 20588608 ( 98%) 806 ( 0%) -------------------- ------------------41176896 ( 98%) 1736 ( 0%) 20969408 (100%) 62 ( 0%) 20969408 (100%) 62 ( 0%) -------------------- ------------------41938816 (100%) 124 ( 0%) ==================== =================== 83115712 ( 99%) 1860 ( 0%) 41176896 ( 98%) 1736 ( 0%) ==================== =================== 83115712 ( 99%) 1860 ( 0%)

yes yes

Inode Information ----------------Number of used inodes: Number of free inodes: Number of allocated inodes: Maximum number of inodes:

4044 78132 82176 82176

Create a file in fileset1 called bigfile1.dat

# dd if=/dev/zero of=/gpfs/fileset1/bigfile1.dat bs=64k count=1000

Record the free space (bigfile1.dat) 5 Create a file in fileset5 called bigfile2
# dd if=/dev/zero of=/gpfs/fileset5/bigfile2 bs=64k count=1000

Record the free space (bigfile2) 6 Questions: Where did the data go for each file? Bigfile1 ______________ Bigfile1.dat ______________ Bigfile2 ______________ Why? Create a couple more files (These will be used in the next step)
# dd if=/dev/zero of=/gpfs/fileset3/bigfile3 bs=64k count=10000 # dd if=/dev/zero of=/gpfs/fileset4/bigfile4 bs=64k count=10000

Step 5: File Management Policies Now that you have data in the file system you are going to manage the placement of the file data using file management policies. For this example your business rules say that all file names that

25

start with the letters "big" need to be moved to pool1. In addition, all files that end in ".dat" should be deleted. 1 To begin, create a policy file /gpfs-course/data/managementpolicy.txt that implements the business rules.
RULE 'datfiles' DELETE WHERE UPPER(name) like '%.DAT' RULE 'bigfiles' MIGRATE TO POOL 'pool1' WHERE UPPER(name) like 'BIG%'

Test the rule set using the mmapplypolicy command

# mmapplypolicy fs1 \-P managementpolicy.txt \-I test

This command will show you what mmapplypolicy will do but will not actually perform the delete or migrate. 3 Actually perform the migration and deletion using the mmapplypolicy command
# mmapplypolicy fs1 \-P managementpolicy.txt

Review: Review the output of the mmapplypolicy command to answer these questions. How many files were deleted? ____________ How many files were moved? ____________ How many KB total were moved? ___________

Step 6: Using External Pools In this step you will use the external pool interface to generate a report. 1 Create two files : expool1.bash, listrule1.txt: File /tmp/expool1.bash
#!/bin/bash dt=`date +%h%d%y-%H_%M_%S` results=/tmp/FileReport_${dt} echo one $1 if [[ $1 == 'MIGRATE' ]];then echo Filelist echo There are `cat $2 | wc -l` files that match >> ${results} cat $2 >> ${results} echo ---echo - The file list report has been placed in ${results} echo ---fi

File listrule1.txt
RULE EXTERNAL POOL 'externalpoolA' EXEC '/tmp/expool1.bash' RULE 'MigToExt' MIGRATE TO POOL 'externalpoolA' WHERE FILE_SIZE > 2

26

Note: You may need to modify the where clause to get a list of files on your file system. 2 Make the external pool script executable
# chmod +x /tmp/expool1.bash

Execute the job

# mmapplypolicy fs1 -P listrule1.txt

It will print output to the screen. When it is done it will print the location of the results file. For example:
The file list report has been placed in /tmp/FileReport_Jul3108-20_15_50

What information do you see in the file?

27

GPFS Exercise Lab 3 Using Replication

In this lab we will configure a file system for replication. Objectives: Enable data and metadata replication Verify and monitor a file's replication status

Requirements: 1. Complete Exercise 1: Installing the cluster 2. List of available devices /dev/sd__ /dev/sd__ /dev/sd__ /dev/sd__ Step 1: Enabling Replication 1 The max replication factor should have been set in Lab 1. Use the mmlsfs command to verify the file system is enabled for replication. A file system is enabled for replication when the maximum number of data and metadata replicas is set to 2. If these parameters are not set to 2 you will need to recreate the file system. To recreate the file system a. Umount the file system b. Delete the file system c. Create the file system and specify -M 2 and -R 2
# mmcrfs /gpfs fs1 \-F pooldesc.txt \-B 64k \-M 2 \-R 2

Where pooldesc.txt is the disk descriptor file from Lab 1 Step 2: Change the failure group on the NSDs 1 View the current disk usage using the mmdf command
# mmdf fs1

Check the status of replication with the mmlsdisk command:

# mmlsdisk fs1

28

The failure group should be set to a value of -1 3 Change the failure group to 1 for nsd1 and nsd3 and to 2 for nsd2 and nsd4 using the mmchdisk command.
# # # # mmchdisk mmchdisk mmchdisk mmchdisk fs1 fs1 fs1 fs1 change change change change -d -d -d -d "nsd1:::dataAndMetadata:1:::" "nsd2:::dataAndMetadata:2:::" "nsd3:::dataOnly:1:::" "nsd4:::dataOnly:2:::"

Verify the changes using the mmdf and mmlsdisk commands:

# mmlsdisk fs1 # mmdf fs1

Notice that data was not written because the default replication level is still set to 1. Now that there are two failure groups you can see how to change the replication status of a file. Step 3: Replicate a file Replication status can bet set at the file level. In this step we will replicate the data and metadata of a single file in the file system. 1 Create a file in the GPFS file system, /gpfs, called bigfile10
# dd if=/dev/zero of=/gpfs/fileset1/bigfile10 bs=64k count=1000

Use the mmlsattr command to check the replication status of the file bigfile10
# mmlsattr /gpfs/fileset1/bigfile10 replication factors metadata(max) data(max) file [flags] ------------- --------- --------------1 ( 2) 1 ( 2) /gpfs/fileset1/bigfile10

Change the file replication status of bigfile10 so that it is replicated in two failure groups using the mmchattr command.
# mmchattr \-m 2 \-r 2 /gpfs/fileset1/bigfile10

Notice that this command take a few moments to execute, as you change the replication status of a file the data is copied before the command completes unless you use the "-I defer" option. 4 Again use the mmlsattr command to check the replication status of the file bigfile10
# mmlsattr /gpfs/fileset1/bigfile10

29

Did you see a change in the replication status of the file? Step 4: Replicate all data in the file system If desired you can replicate all of the data in the file system. In this step we will change the default replication status for the whole file system. 1 Create a file in fileset1 called bigfile11:
# dd if=/dev/zero of=/gpfs/fileset1/bigfile11 bs=64k count=1000

Use the mmlsattr command to check the replication status of the file bigfile11:
# mmlsattr /gpfs/fileset1/bigfile11

Using the mmchfs command change the default replicaton status for fs1:
# mmchfs fs1 \-m 2 \-r 2

Use the mmlsattr command to check the replication status of the file bigfile11:
# mmlsattr /gpfs/fileset1/bigfile11

Has the replication status of bigfile11 changed? _________________ 5 The replication status of existing files does not change until mmrestripefs is run. New files created after the mmchfs command above will get replicated, however. To test this create a new file called bigfile12:
# dd if=/dev/zero of=/gpfs/fileset1/bigfile12 bs=64k count=1000

Use the mmlsattr command to check the replication status of the file bigfile12:
# mmlsattr /gpfs/fileset1/bigfile12

Is the file replicated? 7 You can replicate the existing files in the file system using the mmrestripefs command:
# mmrestripefs fs1 \-R

Recheck the file replication status of bigfile11:

# mmlsattr /gpfs/fileset1/bigfile11

Is the file replicated now?

30

GPFS Exercise Lab 4 Snapshots

In this lab we will use the snapshot feature to create online copies of files. Objectives: Create a file system snapshot Restore a user deleted file from a snapshot image Manage multiple snapshot images

Requirements: 1. Complete Exercise 1: Installing the cluster 2. A File System - Use Exercise 2 to create a file system if you do not already have one. Step 1: Use a snapshot to backup a file A snapshot is a point in time view of a file system. To see how snapshots operate you will create a file, take a snapshot, delete the file then restore the snapshot from the file. 1 Create a file for testing in the /gpfs/fileset1 directory:
# echo "hello world:snap1" > /gpfs/fileset1/snapfile1

Create a snapshot image using the mmcrsnapshot command:

# mmcrsnapshot fs1 snap1 Writing dirty data to disk Quiescing all file system operations Writing dirty data to disk again Creating snapshot. Resuming operations.

Modify the file for testing in the /gpfs/fileset1 directory:

# echo "hello world:snap2" >> /gpfs/fileset1/snapfile1

Create a seconds snapshot image using the mmcrsnapshot command:

# mmcrsnapshot fs1 snap2

View the list of snapshots created using the mmlssnapshot command Example:

31

# mmlssnapshot fs1 gpfs1:~ # mmlssnapshot fs1 Snapshots in file system fs1: Directory SnapId Status snap1 2 Valid snap2 3 Valid

Created Wed Sep 26 11:03:52 2007 Wed Sep 26 11:04:52 2007

6 7

Delete the file /gpfs/fileset1/snapfile1 Now that the file is deleted let's see what is in the snapshots. Take a look at the snapshot images. To view the image change directories to the .snapshot directory
# cd /gpfs/.snapshots

What directories do you see? _____________________ 8 Compare the snapfile1 stored in each snapshot:
# cat snap1/fileset1/snapfile1 [output omitted] # cat snap2/fileset1/snapfile1 [output omitted]

Are the file contents the same? _______________ 9 To restore the file from the snapshot copy the file back into the original location:
# cp /gpfs/.snapshots/snap2/fileset1/snapfile1 \ /gpfs/fileset1/snapfile1

10

When you are done with a snapshot you can delete the snapshot. Delete both of these snapshots using the mmdelsnapshot command:
# mmdelsnapshot fs1 snap1 # mmdelsnapshot fs1 snap2

11

Verify the snapshots were deleted using the mmlssnapshot command:

# mmlssnapshot fs1

32

GPFS Exercise Lab 5 Dynamically Adding a Disk to an Online File System

Objectives: Add a disk to a storage pool online Re-balance existing data in the file system

Requirements: 1. Complete Exercise 1: Installing the cluster 2. A File System (Use Exercise 2 to create a file system if you do not already have one). 3. Device to add /dev/sd___ Step 1: Add a disk to the existing file system 1 Verify that GPFS is running and the file system is mounted using the mmgetstate command and the df command a. The mmgetstate command will show the status of the nodes in the cluster.
# mmgetstate -a

b. The df command will display the mounted GPFS file system.

# df

Create a disk descriptor file /gpfs-course/data/adddisk.txt for the new disk using the format:
#DiskName:serverlist::DiskUsage:FailureGroup:DesiredName:StoragePool /dev/sd_:::dataOnly::nsd5:pool1

Use the mmcrnsd command to create the NSD:

# mmcrnsd -F /gpfs-course/data/adddisk.txt

Verify the disk has been created using the mmlsnsd command:
# mmlsnsd

The disk you just added should show as a (free disk) 5 Add the new NSD to the fs1 file system using the mmadddisk command:

33

# mmadddisk fs1 -F /gpfs-course/data/adddisk.txt

Verify the NSD was added using the mmdf command

# mmdf fs1 disk disk size failure holds holds free KB free KB name in KB group metadata data in full blocks in fragments --------------- ------------- -------- -------- ----- -------------------- ------------------Disks in storage pool: system nsd1 20971520 1 yes yes 20873984 (100%) 284 ( 0%) nsd2 20971520 2 yes yes 20873984 (100%) 202 ( 0%) -------------------------------- ------------------(pool total) 41943040 41747968 (100%) 486 ( 0%) Disks in storage pool: pool1 nsd3 20971520 nsd4 20971520 nsd5 20971520 ------------(pool total) 62914560 (data) (metadata) (total) ============= 104857600 41943040 ============= 104857600 1 no 2 no -1 no yes yes yes 20969408 (100%) 62 ( 0%) 20969408 (100%) 62 ( 0%) 20969408 (100%) 62 ( 0%) -------------------- ------------------62908224 (100%) 186 ( 0%) ==================== =================== 104656192 (100%) 672 ( 0%) 41747968 (100%) 486 ( 0%) ==================== =================== 104656192 (100%) 672 ( 0%)

Inode Information ----------------Number of used inodes: Number of free inodes: Number of allocated inodes: Maximum number of inodes:

4045 78131 82176 82176

Step 2: Re-balancing the data In some cases you may wish to have GPFS re-balance existing data over the new disks that were added to the file system. Often it is not necessary to manually re-balance the data across the new disks. New data that is added to the file system is correctly striped. Re-striping a large file system requires a large number of insert and delete operations and may affect system performance. Plan to perform this task when system demand is low. 1 To re-balance the existing data in the file system use the mmrestripefs command:
# mmrestripefs fs1 b

Use the mmdf command to view the utilization of each disk.

34

GPFS Exercise Lab 6 Permanently Remove GPFS

WARNING!: Performing these steps will remove GPFS and destroy all data on the GPFS file system. Objectives: Remove all traces of GPFS from hosts in a cluster

Requirements: 1. A cluster with multiple nodes of GPFS Step 1: Shut down the necessary services 1 Unmount all file systems
# mmumount all -a

Delete the file system

# mmdelfs fs1

(Replace fs1 with your file system name. Repeat for additional file systems.) 3 Remove all NSDs:
# mmdelnsd nsd1

(Replace NSD1 with your NSD name. Repeat for additional NSDs.) 4 Stop GPFS:
# mmshutdown -a

Step 2: Remove GPFS packages This section assumes GPFS is installed on Linux using the .rpm packages. 1 Verify the packages installed on the current host:
# rpm -qa | grep gpfs

Write down each package name (less the version).

35

Remove each package. For each package listed in step 1, perform the following:
# rpm -e <package-name>

(where package-name> is the name of the package without the version number listed in step 1.) Step 3: Remove Residual Traces of GPFS This section assumes GPFS is installed on Linux using the .rpm packages. 1 Remove gpfs-specific directories and files: # rm -rf /var/mmfs # rm -rf /usr/lpp/mmfs # rm -rf /var/adm/ras/mm* # rm -rf /tmp/mmfs 2 3 Remove the /gpfs mount entry from /etc/fstab. If you appended the GPFS binary directory (/usr/lpp/mmfs/bin) to root's PATH environment variable, remove it.

36