instructions client machine installing client functionality client machine This chapter describes how to install AFS client machines after you have installed the first AFS machine. Some parts of the installation differ depending on whether or not the new client is of the same AFS system type (uses the same AFS binaries) as a previously installed client machine. overview installing client machine Incorporate AFS into the machine's kernel Define the machine's cell membership Define cache location and size Create the /usr/vice/etc/CellServDB file, which determines which foreign cells the client can access in addition to the local cell Create the /afs directory and start the Cache Manager Create and mount volumes for housing AFS client binaries (necessary only for clients of a new system type) Create a link from the local /usr/afsws directory to the AFS directory housing the AFS client binaries Modify the machine's authentication system to enable AFS users to obtain tokens at login Binary Distribution creating /tmp/afsdist directory client machine afsdist directory client machine client machine /tmp/afsdist directory creating /tmp/afsdist directory client machine usr/vice/etc directory client machine client machine /usr/vice/etc directory creating /usr/vice/etc directory client machine If you are not installing from a packaged distribution, create the /usr/vice/etc directory on the local disk, to house client binaries and configuration files. Subsequent instructions copy files from the OpenAFS binary distribution into them. Create the /tmp/afsdist directory as a location to uncompress this distribution, if it does not already exist. # mkdir /usr/vice # mkdir /usr/vice/etc # mkdir /tmp/afsdist Every AFS client machine's kernel must incorporate AFS modifications. Some system types use a dynamic kernel loader program, whereas on other system types you build AFS modifications into a static kernel. Some system types support both methods. Also modify the machine's authentication system so that users obtain an AFS token as they log into the local file system. Using AFS is simpler and more convenient for your users if you make the modifications on all client machines. Otherwise, users must perform a two or three step login procedure (login to the local system, obtain Kerberos credentials, and then issue the klog command). For further discussion of AFS authentication, see the chapter in the OpenAFS Administration Guide about cell configuration and administration issues. For convenience, the following sections group the two procedures by system type. Proceed to the appropriate section. Getting Started on AIX Systems Getting Started on HP-UX Systems Getting Started on IRIX Systems Getting Started on Linux Systems Getting Started on Solaris Systems incorporating AFS kernel extensions client machine AIX AFS kernel extensions on client machine AIX client machine AFS kernel extensions on AIX AIX AFS kernel extensions on client machine enabling AFS login client machine AIX AFS login on client machine AIX client machine AFS login on AIX AIX AFS login on client machine secondary authentication system (AIX) client machine In this section you load AFS into the AIX kernel. Then incorporate AFS modifications into the machine's secondary authentication system, if you wish to enable AFS login. The AIX kernel extension facility is the dynamic kernel loader provided by IBM Corporation. AIX does not support incorporation of AFS modifications during a kernel build. For AFS to function correctly, the kernel extension facility must run each time the machine reboots, so the AFS initialization script (included in the AFS distribution) invokes it automatically. In this section you copy the script to the conventional location and edit it to select the appropriate options depending on whether NFS is also to run. After editing the script, you run it to incorporate AFS into the kernel. In a later section you verify that the script correctly initializes the Cache Manager, then configure the AIX inittab file so that the script runs automatically at reboot. Unpack the distribution tarball. The examples below assume that you have unpacked the files into the /tmp/afsdist directory. If you pick a different location, substitute this in all of the following examples. Once you have unpacked the distribution, change directory as indicated. # cd /tmp/afsdist/rs_aix42/root.client/usr/vice/etc Copy the AFS kernel library files to the local /usr/vice/etc/dkload directory, and the AFS initialization script to the /etc directory. # cp -rp dkload /usr/vice/etc # cp -p rc.afs /etc/rc.afs Edit the /etc/rc.afs script, setting the NFS variable as indicated. If the machine is not to function as an NFS/AFS Translator, set the NFS variable as follows. NFS=$NFS_NONE If the machine is to function as an NFS/AFS Translator and is running AIX 4.2.1 or higher, set the NFS variable as follows. Note that NFS must already be loaded into the kernel, which happens automatically on systems running AIX 4.1.1 and later, as long as the file /etc/exports exists. NFS=$NFS_IAUTH Invoke the /etc/rc.afs script to load AFS modifications into the kernel. You can ignore any error messages about the inability to start the BOS Server or the Cache Manager or AFS client. # /etc/rc.afs In modern AFS installations, you should be using Kerberos v5 for user login, and obtaining AFS tokens following this authentication step. There are currently no instructions available on configuring AIX to automatically obtain AFS tokens at login. Following login, users can obtain tokens by running the aklog command Sites which still require kaserver or external Kerberos v4 authentication should consult Enabling kaserver based AFS Login on AIX Systems for details of how to enable AIX login. Proceed to Loading and Creating Client Files. incorporating AFS kernel extensions client machine HP-UX AFS kernel extensions on client machine HP-UX client machine AFS kernel extensions on HP-UX HP-UX AFS-modified kernel on client machine enabling AFS login client machine HP-UX AFS login on client machine HP-UX client machine AFS login on HP-UX HP-UX AFS login on client machine PAM on HP-UX client machine In this section you build AFS into the HP-UX kernel. Then incorporate AFS modifications into the machine's Pluggable Authentication Module (PAM) system, if you wish to enable AFS login. On HP-UX systems, you must build AFS modifications into a new static kernel; HP-UX does not support dynamic loading. If the machine's hardware and software configuration exactly matches another HP-UX machine on which AFS is already built into the kernel, you can choose to copy the kernel from that machine to this one. In general, however, it is better to build AFS modifications into the kernel on each machine according to the following instructions. Move the existing kernel-related files to a safe location. # cp /stand/vmunix /stand/vmunix.noafs # cp /stand/system /stand/system.noafs Unpack the OpenAFS HP-UX distribution tarball. The examples below assume that you have unpacked the files into the /tmp/afsdist directory. If you pick a different location, substitute this in all of the following examples. Once you have unpacked the distribution, change directory as indicated. # cd /tmp/afsdist/hp_ux110/root.client Copy the AFS initialization file to the local directory for initialization files (by convention, /sbin/init.d on HP-UX machines). Note the removal of the .rc extension as you copy the file. # cp usr/vice/etc/afs.rc /sbin/init.d/afs Copy the file afs.driver to the local /usr/conf/master.d directory, changing its name to afs as you do. # cp usr/vice/etc/afs.driver /usr/conf/master.d/afs Copy the AFS kernel module to the local /usr/conf/lib directory. If the machine's kernel supports NFS server functionality: # cp bin/libafs.a /usr/conf/lib If the machine's kernel does not support NFS server functionality, change the file's name as you copy it: # cp bin/libafs.nonfs.a /usr/conf/lib/libafs.a Incorporate the AFS driver into the kernel, either using the SAM program or a series of individual commands. To use the SAM program: Invoke the SAM program, specifying the hostname of the local machine as local_hostname. The SAM graphical user interface pops up. # sam -display local_hostname:0 Choose the Kernel Configuration icon, then the Drivers icon. From the list of drivers, select afs. Open the pull-down Actions menu and choose the Add Driver to Kernel option. Open the Actions menu again and choose the Create a New Kernel option. Confirm your choices by choosing Yes and OK when prompted by subsequent pop-up windows. The SAM program builds the kernel and reboots the system. Login again as the superuser root. login: root Password: root_password To use individual commands: Edit the file /stand/system, adding an entry for afs to the Subsystems section. Change to the /stand/build directory and issue the mk_kernel command to build the kernel. # cd /stand/build # mk_kernel Move the new kernel to the standard location (/stand/vmunix), reboot the machine to start using it, and login again as the superuser root. # mv /stand/build/vmunix_test /stand/vmunix # cd / # shutdown -r now login: root Password: root_password At this point you incorporate AFS into the operating system's Pluggable Authentication Module (PAM) scheme. PAM integrates all authentication mechanisms on the machine, including login, to provide the security infrastructure for authenticated access to and from the machine. In modern AFS installations, you should be using Kerberos v5 for user login, and obtaining AFS tokens subsequent to this authentication step. OpenAFS does not currently distribute a PAM module allowing AFS tokens to be automatically gained at login. Whilst there are a number of third party modules providing this functionality, it is not know if these have been tested with HP/UX. Following login, users can obtain tokens by running the aklog command If you are at a site which still requires kaserver or external Kerberos v4 based authentication, please consult Enabling kaserver based AFS Login on HP-UX systems for further installation instructions. Proceed to Loading and Creating Client Files. incorporating AFS kernel extensions client machine IRIX AFS kernel extensions on client machine IRIX client machine AFS kernel extensions on IRIX In this section you incorporate AFS into the IRIX kernel, choosing one of two methods: Dynamic loading using the ml program distributed by Silicon Graphics, Incorporated (SGI). Building a new static kernel. Then see Enabling AFS Login on IRIX Systems to read about integrated AFS login on IRIX systems. In preparation for either dynamic loading or kernel building, perform the following procedures: Unpack the OpenAFS IRIX distribution tarball. The examples below assume that you have unpacked the files into the /tmp/afsdist directory. If you pick a different location, substitue this in all of the following examples. Once you have unpacked the distribution, change directory as indicated. # cd /tmp/afsdist/sgi_65/root.client Copy the AFS initialization script to the local directory for initialization files (by convention, /etc/init.d on IRIX machines). Note the removal of the .rc extension as you copy the script. # cp -p usr/vice/etc/afs.rc /etc/init.d/afs Issue the uname -m command to determine the machine's CPU board type. The IPxx value in the output must match one of the supported CPU board types listed in the OpenAFS Release Notes for the current version of AFS. # uname -m Proceed to either Loading AFS into the IRIX Kernel or Building AFS into the IRIX Kernel. IRIX AFS kernel extensions on client machine afsml variable (IRIX) client machine variables afsml (IRIX) client machine IRIX afsml variable client machine afsxnfs variable (IRIX) client machine variables afsxnfs (IRIX) client machine IRIX afsxnfs variable client machine The ml program is the dynamic kernel loader provided by SGI for IRIX systems. If you use it rather than building AFS modifications into a static kernel, then for AFS to function correctly the ml program must run each time the machine reboots. Therefore, the AFS initialization script (included in the OpenAFS Binary Distribution) invokes it automatically when the afsml configuration variable is activated. In this section you activate the variable and run the script. In a later section you verify that the script correctly initializes the Cache Manager, then create the links that incorporate AFS into the IRIX startup and shutdown sequence. Create the local /usr/vice/etc/sgiload directory to house the AFS kernel library file. # mkdir /usr/vice/etc/sgiload Copy the appropriate AFS kernel library file to the /usr/vice/etc/sgiload directory. The IPxx portion of the library file name must match the value previously returned by the uname -m command. Also choose the file appropriate to whether the machine's kernel supports NFS server functionality (NFS must be supported for the machine to act as an NFS/AFS Translator). Single- and multiprocessor machines use the same library file. (You can choose to copy all of the kernel library files into the /usr/vice/etc/sgiload directory, but they require a significant amount of space.) If the machine's kernel supports NFS server functionality: # cp -p usr/vice/etc/sgiload/libafs.IPxx.o /usr/vice/etc/sgiload If the machine's kernel does not support NFS server functionality: # cp -p usr/vice/etc/sgiload/libafs.IPxx.nonfs.o \ /usr/vice/etc/sgiload Issue the chkconfig command to activate the afsml configuration variable. # /etc/chkconfig -f afsml on If the machine is to function as an NFS/AFS Translator and the kernel supports NFS server functionality, activate the afsxnfs variable. # /etc/chkconfig -f afsxnfs on Run the /etc/init.d/afs script to load AFS extensions into the kernel. The script invokes the ml command, automatically determining which kernel library file to use based on this machine's CPU type and the activation state of the afsxnfs variable. You can ignore any error messages about the inability to start the BOS Server or the Cache Manager or AFS client. # /etc/init.d/afs start Proceed to Enabling AFS Login on IRIX Systems. IRIX AFS-modified kernel on client machine If you prefer to build a kernel, and the machine's hardware and software configuration exactly matches another IRIX machine on which AFS is already built into the kernel, you can choose to copy the kernel from that machine to this one. In general, however, it is better to build AFS modifications into the kernel on each machine according to the following instructions. Copy the kernel initialization file afs.sm to the local /var/sysgen/system directory, and the kernel master file afs to the local /var/sysgen/master.d directory. # cp -p bin/afs.sm /var/sysgen/system # cp -p bin/afs /var/sysgen/master.d Copy the appropriate AFS kernel library file to the local file /var/sysgen/boot/afs.a; the IPxx portion of the library file name must match the value previously returned by the uname -m command. Also choose the file appropriate to whether the machine's kernel supports NFS server functionality (NFS must be supported for the machine to act as an NFS/AFS Translator). Single- and multiprocessor machines use the same library file. If the machine's kernel supports NFS server functionality: # cp -p bin/libafs.IPxx.a /var/sysgen/boot/afs.a If the machine's kernel does not support NFS server functionality: # cp -p bin/libafs.IPxx.nonfs.a /var/sysgen/boot/afs.a Issue the chkconfig command to deactivate the afsml configuration variable. # /etc/chkconfig -f afsml off If the machine is to function as an NFS/AFS Translator and the kernel supports NFS server functionality, activate the afsxnfs variable. # /etc/chkconfig -f afsxnfs on Copy the existing kernel file, /unix, to a safe location. Compile the new kernel, which is created in the file /unix.install. It overwrites the existing /unix file when the machine reboots in the next step. # cp /unix /unix_noafs # autoconfig Reboot the machine to start using the new kernel, and login again as the superuser root. # cd / # shutdown -i6 -g0 -y login: root Password: root_password Proceed to Enabling AFS Login on IRIX Systems. enabling AFS login client machine IRIX AFS login on client machine IRIX client machine AFS login on IRIX Whilst the standard IRIX command-line login program and the graphical xdm login program both have the ability to grant AFS tokens, this ability relies upon the deprecated kaserver authentication system. As this system is not recommended for new installations, this is not documented here. Users who have been successfully authenticated via Kerberos 5 authentication may obtain AFS tokens following login by running the aklog command. If you are at a site which still requires kaserver or external Kerberos v4 based authentication, please consult Enabling kaserver based AFS Login on Linux Systems for further installation instructions. Proceed to Loading and Creating Client Files. incorporating AFS kernel extensions client machine Linux AFS kernel extensions on client machine Linux client machine AFS kernel extensions on Linux Linux AFS kernel extensions on client machine enabling AFS login client machine Linux AFS login on client machine Linux client machine AFS login on Linux Linux AFS login on client machine PAM on Linux client machine In this section you load AFS into the Linux kernel. Then incorporate AFS modifications into the machine's Pluggable Authentication Module (PAM) system, if you wish to enable AFS login. The modprobe program is the dynamic kernel loader for Linux. Linux does not support incorporation of AFS modifications during a kernel build. For AFS to function correctly, the modprobe program must run each time the machine reboots, so your distributions's AFS initialization script invokes it automatically. The script also includes commands that select the appropriate AFS library file automatically. In this section you run the script. In a later section you also verify that the script correctly initializes the Cache Manager, then activate a configuration variable, which results in the script being incorporated into the Linux startup and shutdown sequence. The procedure for starting up OpenAFS depends upon your distribution OpenAFS ships RPMS for all current Fedora and RHEL releases. Download and install the RPM set for your operating system. RPMs are available from the OpenAFS web site. You will need the openafs and openfs-client packages, along with an openafs-kernel package matching your current, running ,kernel. You can find the version of your current kernel by running # uname -r 2.6.20-1.2933.fc6 Once downloaded, the packages may be installed with the rpm command # rpm -U openafs-* openafs-client-* openafs-server-* openafs-kernel-* If you are running a system where the OpenAFS Binary Distribution is provided as a tar file, or where you have built the system from source yourself, you need to install the relevant components by hand Unpack the distribution tarball. The examples below assume that you have unpacked the files into the /tmp/afsdistdirectory. If you pick a different location, substitute this in all of the following examples. Once you have unpacked the distribution, change directory as indicated. # cd /tmp/afsdist/linux/root.client/usr/vice/etc Copy the AFS kernel library files to the local /usr/vice/etc/modload directory. The filenames for the libraries have the format libafs-version.o, where version indicates the kernel build level. The string .mp in the version indicates that the file is appropriate for machines running a multiprocessor kernel. # cp -rp modload /usr/vice/etc Copy the AFS initialization script to the local directory for initialization files (by convention, /etc/rc.d/init.d on Linux machines). Note the removal of the .rc extension as you copy the script. # cp -p afs.rc /etc/rc.d/init.d/afs At this point you incorporate AFS into the operating system's Pluggable Authentication Module (PAM) scheme. PAM integrates all authentication mechanisms on the machine, including login, to provide the security infrastructure for authenticated access to and from the machine. At this time, we recommend that new sites requiring AFS credentials to be gained as part of PAM authentication use Russ Alberry's pam_afs_session, rather than utilising the bundled pam_afs2 module. A typical PAM stack should authenticate the user using an external Kerberos V service, and then use the AFS PAM module to obtain AFS credentials in the session section If you are at a site which still requires kaserver or external Kerberos v4 based authentication, please consult Enabling kaserver based AFS Login on Linux Systems for further installation instructions. Proceed to Loading and Creating Client Files. incorporating AFS kernel extensions client machine Solaris AFS kernel extensions on client machine Solaris client machine AFS kernel extensions on Solaris Solaris AFS kernel extensions on client machine enabling AFS login client machine Solaris AFS login on client machine Solaris client machine AFS login on Solaris Solaris AFS login on client machine PAM on Solaris client machine Solaris file systems clean-up script on client machine file systems clean-up script (Solaris) client machine scripts file systems clean-up (Solaris) client machine In this section you load AFS into the Solaris kernel. Then incorporate AFS modifications into the machine's Pluggable Authentication Module (PAM) system, if you wish to enable AFS login. The modload program is the dynamic kernel loader provided by Sun Microsystems for Solaris systems. Solaris does not support incorporation of AFS modifications during a kernel build. For AFS to function correctly, the modload program must run each time the machine reboots, so the AFS initialization script (included on the AFS CD-ROM) invokes it automatically. In this section you copy the appropriate AFS library file to the location where the modload program accesses it and then run the script. In a later section you verify that the script correctly initializes the Cache Manager, then create the links that incorporate AFS into the Solaris startup and shutdown sequence. Unpack the OpenAFS Solaris distribution tarball. The examples below assume that you have unpacked the files into the /tmp/afsdist directory. If you pick a diferent location, substitute this in all of the following exmaples. Once you have unpacked the distribution, change directory as indicated. # cd /tmp/afsdist/sun4x_56/root.client/usr/vice/etc Copy the AFS initialization script to the local directory for initialization files (by convention, /etc/init.d on Solaris machines). Note the removal of the .rc extension as you copy the script. # cp -p afs.rc /etc/init.d/afs Copy the appropriate AFS kernel library file to the local file /kernel/fs/afs. If the machine is running Solaris 2.6 or the 32-bit version of Solaris 7, its kernel supports NFS server functionality, and the nfsd process is running: # cp -p modload/libafs.o /kernel/fs/afs If the machine is running Solaris 2.6 or the 32-bit version of Solaris 7, and its kernel does not support NFS server functionality or the nfsd process is not running: # cp -p modload/libafs.nonfs.o /kernel/fs/afs If the machine is running the 64-bit version of Solaris 7, its kernel supports NFS server functionality, and the nfsd process is running: # cp -p modload/libafs64.o /kernel/fs/sparcv9/afs If the machine is running the 64-bit version of Solaris 7, and its kernel does not support NFS server functionality or the nfsd process is not running: # cp -p modload/libafs64.nonfs.o /kernel/fs/sparcv9/afs Run the AFS initialization script to load AFS modifications into the kernel. You can ignore any error messages about the inability to start the BOS Server or the Cache Manager or AFS client. # /etc/init.d/afs start When an entry called afs does not already exist in the local /etc/name_to_sysnum file, the script automatically creates it and reboots the machine to start using the new version of the file. If this happens, log in again as the superuser root after the reboot and run the initialization script again. This time the required entry exists in the /etc/name_to_sysnum file, and the modload program runs. login: root Password: root_password # /etc/init.d/afs start At this point you incorporate AFS into the operating system's Pluggable Authentication Module (PAM) scheme. PAM integrates all authentication mechanisms on the machine, including login, to provide the security infrastructure for authenticated access to and from the machine. In modern AFS installations, you should be using Kerberos v5 for user login, and obtaining AFS tokens subsequent to this authentication step. OpenAFS does not currently distribute a PAM module allowing AFS tokens to be automatically gained at login. Some of these, such as pam-krb5 and pam-afs-session from http://www.eyrie.org/~eagle/software/ or pam_afs2 from ftp://achilles.ctd.anl.gov/pub/DEE/pam_afs2-0.1.tar, have been tested with Solaris. If you are at a site which still requires kaserver or external Kerberos v4 based authentication, please consult Enabling kaserver based AFS Login on Solaris Systems for further installation instructions. Some Solaris distributions include a script that locates and removes unneeded files from various file systems. Its conventional location is /usr/lib/fs/nfs/nfsfind. The script generally uses an argument to the find command to define which file systems to search. In this step you modify the command to exclude the /afs directory. Otherwise, the command traverses the AFS filespace of every cell that is accessible from the machine, which can take many hours. The following alterations are possibilities, but you must verify that they are appropriate for your cell. The first possible alteration is to add the -local flag to the existing command, so that it looks like the following: find $dir -local -name .nfs\* -mtime +7 -mount -exec rm -f {} \; Another alternative is to exclude any directories whose names begin with the lowercase letter a or a non-alphabetic character. find /[A-Zb-z]* remainder of existing command Do not use the following command, which still searches under the /afs directory, looking for a subdirectory of type 4.2. find / -fstype 4.2 /* do not use */ Proceed to Loading and Creating Client Files. Binary Distribution copying client files from client machine client machine copying client files to local disk copying client files to local disk client machine cell name setting in client ThisCell file client machine setting cell name in client ThisCell file client machine client machine cell membership client machine ThisCell file ThisCell file (client) client machine CellServDB file (client) creating on client machine database server machine entry in client CellServDB file on client machine creating CellServDB file (client) client machine client machine CellServDB file creating during initial installation If you are using a non-packaged distribution (that is, one provided as a tarball) you should now copy files from the istribution to the /usr/vice/etc directory. On some platforms that use a dynamic loader program to incorporate AFS modifications into the kernel, you have already copied over some the files. Copying them again does no harm. Every AFS client machine has a copy of the /usr/vice/etc/ThisCell file on its local disk to define the machine's cell membership for the AFS client programs that run on it. Among other functions, this file determines the following: The cell in which users authenticate when they log onto the machine, assuming it is using an AFS-modified login utility The cell in which users authenticate by default when they issue the aklog command The cell membership of the AFS server processes that the AFS command interpreters on this machine contact by default Similarly, the /usr/vice/etc/CellServDB file on a client machine's local disk lists the database server machines in each cell that the local Cache Manager can contact. If there is no entry in the file for a cell, or the list of database server machines is wrong, then users working on this machine cannot access the cell. The chapter in the OpenAFS Administration Guide about administering client machines explains how to maintain the file after creating it. A version of the client CellServDB file was created during the installation of your cell's first machine (in Creating the Client CellServDB File). It is probably also appropriate for use on this machine. Remember that the Cache Manager consults the /usr/vice/etc/CellServDB file only at reboot, when it copies the information into the kernel. For the Cache Manager to perform properly, the CellServDB file must be accurate at all times. Refer to the chapter in the OpenAFS Administration Guide about administering client machines for instructions on updating this file, with or without rebooting. If you have not already done so, unpack the distribution tarball for this machine's system type into a suitable location on the filesystem, such as /tmp/afsdist. If you use a different location, substitue that in the examples that follow. Copy files to the local /usr/vice/etc directory. This step places a copy of the AFS initialization script (and related files, if applicable) into the /usr/vice/etc directory. In the preceding instructions for incorporating AFS into the kernel, you copied the script directly to the operating system's conventional location for initialization files. When you incorporate AFS into the machine's startup sequence in a later step, you can choose to link the two files. On some system types that use a dynamic kernel loader program, you previously copied AFS library files into a subdirectory of the /usr/vice/etc directory. On other system types, you copied the appropriate AFS library file directly to the directory where the operating system accesses it. The following commands do not copy or recopy the AFS library files into the /usr/vice/etc directory, because on some system types the library files consume a large amount of space. If you want to copy them, add the -r flag to the first cp command and skip the second cp command. # cd /cdrom/sysname/root.client/usr/vice/etc # cp -p * /usr/vice/etc # cp -rp C /usr/vice/etc Create the /usr/vice/etc/ThisCell file. # echo "cellname" > /usr/vice/etc/ThisCell Create the /usr/vice/etc/CellServDB file. Use a network file transfer program such as sftp or scp to copy it from one of the following sources, which are listed in decreasing order of preference: Your cell's central CellServDB source file (the conventional location is /afs/cellname/common/etc/CellServDB) The global CellServDB file maintained at grand.central.org An existing client machine in your cell The CellServDB.sample file included in the sysname/root.client/usr/vice/etc directory of each OpenAFS distribution; add an entry for the local cell by following the instructions in Creating the Client CellServDB File client cache cache AFS cache cache disk cache cache memory cache cache cache requirements cache choosing size requirements cache usr/vice/etc/cacheinfo cacheinfo file cacheinfo file files cacheinfo usr/vice/cache directory directories /usr/vice/cache cache configuring client machine configuring cache client machine setting cache size and location client machine client machine cache size and location The Cache Manager uses a cache on the local disk or in machine memory to store local copies of files fetched from file server machines. As the afsd program initializes the Cache Manager, it sets basic cache configuration parameters according to definitions in the local /usr/vice/etc/cacheinfo file. The file has three fields: The first field names the local directory on which to mount the AFS filespace. The conventional location is the /afs directory. The second field defines the local disk directory to use for the disk cache. The conventional location is the /usr/vice/cache directory, but you can specify an alternate directory if another partition has more space available. There must always be a value in this field, but the Cache Manager ignores it if the machine uses a memory cache. The third field specifies the number of kilobyte (1024 byte) blocks to allocate for the cache. The values you define must meet the following requirements. On a machine using a disk cache, the Cache Manager expects always to be able to use the amount of space specified in the third field. Failure to meet this requirement can cause serious problems, some of which can be repaired only by rebooting. You must prevent non-AFS processes from filling up the cache partition. The simplest way is to devote a partition to the cache exclusively. The amount of space available in memory or on the partition housing the disk cache directory imposes an absolute limit on cache size. The maximum supported cache size can vary in each AFS release; see the OpenAFS Release Notes for the current version. For a disk cache, you cannot specify a value in the third field that exceeds 95% of the space available on the partition mounted at the directory named in the second field. If you violate this restriction, the afsd program exits without starting the Cache Manager and prints an appropriate message on the standard output stream. A value of 90% is more appropriate on most machines. Some operating systems (such as AIX) do not automatically reserve some space to prevent the partition from filling completely; for them, a smaller value (say, 80% to 85% of the space available) is more appropriate. For a memory cache, you must leave enough memory for other processes and applications to run. If you try to allocate more memory than is actually available, the afsd program exits without initializing the Cache Manager and produces the following message on the standard output stream. afsd: memCache allocation failure at number KB The number value is how many kilobytes were allocated just before the failure, and so indicates the approximate amount of memory available. Within these hard limits, the factors that determine appropriate cache size include the number of users working on the machine, the size of the files with which they work, and (for a memory cache) the number of processes that run on the machine. The higher the demand from these factors, the larger the cache needs to be to maintain good performance. Disk caches smaller than 10 MB do not generally perform well. Machines serving multiple users usually perform better with a cache of at least 60 to 70 MB. The point at which enlarging the cache further does not really improve performance depends on the factors mentioned previously and is difficult to predict. Memory caches smaller than 1 MB are nonfunctional, and the performance of caches smaller than 5 MB is usually unsatisfactory. Suitable upper limits are similar to those for disk caches but are probably determined more by the demands on memory from other sources on the machine (number of users and processes). Machines running only a few processes possibly can use a smaller memory cache. Not all file system types that an operating system supports are necessarily supported for use as the cache partition. For possible restrictions, see the OpenAFS Release Notes. To configure the disk cache, perform the following procedures: Create the local directory to use for caching. The following instruction shows the conventional location, /usr/vice/cache. If you are devoting a partition exclusively to caching, as recommended, you must also configure it, make a file system on it, and mount it at the directory created in this step. # mkdir /usr/vice/cache Create the cacheinfo file to define the configuration parameters discussed previously. The following instruction shows the standard mount location, /afs, and the standard cache location, /usr/vice/cache. # echo "/afs:/usr/vice/cache:#blocks" > /usr/vice/etc/cacheinfo The following example defines the disk cache size as 50,000 KB: # echo "/afs:/usr/vice/cache:50000" > /usr/vice/etc/cacheinfo To configure a memory cache, create the cacheinfo file to define the configuration parameters discussed previously. The following instruction shows the standard mount location, /afs, and the standard cache location, /usr/vice/cache (though the exact value of the latter is irrelevant for a memory cache). # echo "/afs:/usr/vice/cache:#blocks" > /usr/vice/etc/cacheinfo The following example allocates 25,000 KB of memory for the cache. # echo "/afs:/usr/vice/cache:25000" > /usr/vice/etc/cacheinfo afs (/afs) directory creating client machine afs (/afs) directory as root of AFS filespace AFS filespace root at /afs directory directories /afs afsd options file (Linux) files afsd options file (Linux) files afs afsd options file (Linux) afs file afsd options file (Linux) etc/sysconfig/afs afs file Linux afsd options file client machine afsd options file (Linux) afsd command in AFS init. script commands afsd OPTIONS variable in AFS initialization file files AFS initialization AFS initialization script scripts AFS initialization AFS initialization script AFS initialization script setting afsd parameters client machine client machine afsd command parameters variables OPTIONS (in AFS initialization file) environment variables variables Cache Manager client machine configuring Cache Manager client machine client machine Cache Manager files vfs (AIX) vfs file etc/vfs file AIX editing /etc/vfs file client machine vfs file (AIX) By convention, the Cache Manager mounts the AFS filespace on the local /afs directory. In this section you create that directory. The afsd program sets several cache configuration parameters as it initializes the Cache Manager, and starts daemons that improve performance. You can use the afsd command's arguments to override the parameters' default values and to change the number of some of the daemons. Depending on the machine's cache size, its amount of RAM, and how many people work on it, you can sometimes improve Cache Manager performance by overriding the default values. For a discussion of all of the afsd command's arguments, see its reference page in the OpenAFS Administration Reference. On platforms using the standard 'afs' initialisation script (this does not apply to Fedora or RHEL based distributions), the afsd command line in the AFS initialization script on each system type includes an OPTIONS variable. You can use it to set nondefault values for the command's arguments, in one of the following ways: You can create an afsd options file that sets values for arguments to the afsd command. If the file exists, its contents are automatically substituted for the OPTIONS variable in the AFS initialization script. The AFS distribution for some system types includes an options file; on other system types, you must create it. You use two variables in the AFS initialization script to specify the path to the options file: CONFIG and AFSDOPT. On system types that define a conventional directory for configuration files, the CONFIG variable indicates it by default; otherwise, the variable indicates an appropriate location. List the desired afsd options on a single line in the options file, separating each option with one or more spaces. The following example sets the -stat argument to 2500, the -daemons argument to 4, and the -volumes argument to 100. -stat 2500 -daemons 4 -volumes 100 On a machine that uses a disk cache, you can set the OPTIONS variable in the AFS initialization script to one of $SMALL, $MEDIUM, or $LARGE. The AFS initialization script uses one of these settings if the afsd options file named by the AFSDOPT variable does not exist. In the script as distributed, the OPTIONS variable is set to the value $MEDIUM. Do not set the OPTIONS variable to $SMALL, $MEDIUM, or $LARGE on a machine that uses a memory cache. The arguments it sets are appropriate only on a machine that uses a disk cache. The script (or on some system types the afsd options file named by the AFSDOPT variable) defines a value for each of SMALL, MEDIUM, and LARGE that sets afsd command arguments appropriately for client machines of different sizes: SMALL is suitable for a small machine that serves one or two users and has approximately 8 MB of RAM and a 20-MB cache MEDIUM is suitable for a medium-sized machine that serves two to six users and has 16 MB of RAM and a 40-MB cache LARGE is suitable for a large machine that serves five to ten users and has 32 MB of RAM and a 100-MB cache You can choose not to create an afsd options file and to set the OPTIONS variable in the initialization script to a null value rather than to the default $MEDIUM value. You can then either set arguments directly on the afsd command line in the script, or set no arguments (and so accept default values for all Cache Manager parameters). If you are running on a Fedora or RHEL based system, the openafs-client initialization script behaves differently from that described above. It sources /etc/sysconfig/openafs, in which the AFSD_ARGS variable may be set to contain any, or all, of the afsd options detailed above. Note that this script does not support setting an OPTIONS variable, or the SMALL, MEDIUM and LARGE methods of defining cache size. Create the local directory on which to mount the AFS filespace, by convention /afs. If the directory already exists, verify that it is empty. # mkdir /afs On AIX systems, add the following line to the /etc/vfs file. It enables AIX to unmount AFS correctly during shutdown. afs 4 none none On non-package based Linux systems, copy the afsd options file from the /usr/vice/etc directory to the /etc/sysconfig directory, removing the .conf extension as you do so. # cp /usr/vice/etc/afs.conf /etc/sysconfig/afs Edit the machine's AFS initialization script or afsd options file to set appropriate values for afsd command parameters. The appropriate file for each system type is as follows: On AIX systems, /etc/rc.afs On HP-UX systems, /sbin/init.d/afs On IRIX systems, /etc/init.d/afs On Fedora and RHEL systems, /etc/sysconfig/openafs On Linux systems, /etc/sysconfig/afs (the afsd options file) On Solaris systems, /etc/init.d/afs Use one of the methods described in the introduction to this section to add the following flags to the afsd command line. Also set any performance-related arguments you wish. Add the -memcache flag if the machine is to use a memory cache. Add the -verbose flag to display a trace of the Cache Manager's initialization on the standard output stream. AFS initialization script running client machine client machine AFS initialization script running AFS init. script client machine installing AFS initialization script client machine AFS initialization script adding to machine startup sequence client machine In this section you run the AFS initialization script to start the Cache Manager. If the script works correctly, perform the steps that incorporate it into the machine's startup and shutdown sequence. If there are problems during the initialization, attempt to resolve them. The AFS Product Support group can provide assistance if necessary. On machines that use a disk cache, it can take a while for the afsd program to run the first time on a machine, because it must create all of the Vn files in the cache directory. Subsequent Cache Manager initializations do not take nearly as long, because the Vn files already exist. On system types that use a dynamic loader program, you must reboot the machine before running the initialization script, so that it can freshly load AFS modifications into the kernel. Proceed to the instructions for your system type: Running the Script on AIX Systems Running the Script on HP-UX Systems Running the Script on IRIX Systems Running the Script on Linux Systems Running the Script on Solaris Systems AIX AFS initialization script on client machine rc.afs file (AFS init. file for AIX) files rc.afs etc/rc.afs rc.afs file Reboot the machine and log in again as the local superuser root. # cd / # shutdown -r now login: root Password: root_password Run the AFS initialization script. # /etc/rc.afs Edit the AIX initialization file, /etc/inittab, adding the following line to invoke the AFS initialization script. Place it just after the line that starts NFS daemons. rcafs:2:wait:/etc/rc.afs > /dev/console 2>&1 # Start AFS services (Optional) There are now copies of the AFS initialization file in both the /usr/vice/etc and /etc directories. If you want to avoid potential confusion by guaranteeing that they are always the same, create a link between them. You can always retrieve the original script from the AFS CD-ROM if necessary. # cd /usr/vice/etc # rm rc.afs # ln -s /etc/rc.afs If a volume for housing AFS binaries for this machine's system type does not already exist, proceed to Setting Up Volumes and Loading Binaries into AFS. Otherwise, the installation is complete. afs file AFS initialization file files afs AFS initialization file HP-UX AFS initialization script on client machine Run the AFS initialization script. # /sbin/init.d/afs start Change to the /sbin/init.d directory and issue the ln -s command to create symbolic links that incorporate the AFS initialization script into the HP-UX startup and shutdown sequence. # cd /sbin/init.d # ln -s ../init.d/afs /sbin/rc2.d/S460afs # ln -s ../init.d/afs /sbin/rc2.d/K800afs (Optional) There are now copies of the AFS initialization file in both the /usr/vice/etc and /sbin/init.d directories. If you want to avoid potential confusion by guaranteeing that they are always the same, create a link between them. You can always retrieve the original script from the AFS CD-ROM if necessary. # cd /usr/vice/etc # rm afs.rc # ln -s /sbin/init.d/afs afs.rc If a volume for housing AFS binaries for this machine's system type does not already exist, proceed to Setting Up Volumes and Loading Binaries into AFS. Otherwise, the installation is complete. afs file AFS initialization file files afs AFS initialization file IRIX AFS initialization script on client machine etc/init.d/afs afs file afsclient variable (IRIX) client machine variables afsclient (IRIX) client machine IRIX afsclient variable client machine If you have configured the machine to use the ml dynamic loader program, reboot the machine and log in again as the local superuser root. # cd / # shutdown -i6 -g0 -y login: root Password: root_password Issue the chkconfig command to activate the afsclient configuration variable. # /etc/chkconfig -f afsclient on Run the AFS initialization script. # /etc/init.d/afs start Change to the /etc/init.d directory and issue the ln -s command to create symbolic links that incorporate the AFS initialization script into the IRIX startup and shutdown sequence. # cd /etc/init.d # ln -s ../init.d/afs /etc/rc2.d/S35afs # ln -s ../init.d/afs /etc/rc0.d/K35afs (Optional) There are now copies of the AFS initialization file in both the /usr/vice/etc and /etc/init.d directories. If you want to avoid potential confusion by guaranteeing that they are always the same, create a link between them. You can always retrieve the original script from the AFS CD-ROM if necessary. # cd /usr/vice/etc # rm afs.rc # ln -s /etc/init.d/afs afs.rc If a volume for housing AFS binaries for this machine's system type does not already exist, proceed to Setting Up Volumes and Loading Binaries into AFS. Otherwise, the installation is complete. afs file AFS initialization file files afs AFS initialization file etc/rc.d/init.d/afs afs file Linux AFS initialization script on client machine Reboot the machine and log in again as the local superuser root. # cd / # shutdown -r now login: root Password: root_password Run the AFS initialization script. # /etc/rc.d/init.d/openafs-client start Issue the chkconfig command to activate the openafs-client configuration variable. Based on the instruction in the AFS initialization file that begins with the string #chkconfig, the command automatically creates the symbolic links that incorporate the script into the Linux startup and shutdown sequence. # /sbin/chkconfig --add openafs-client Reboot the machine and log in again as the local superuser root. # cd / # shutdown -r now login: root Password: root_password Run the AFS initialization script. # /etc/rc.d/init.d/afs start Issue the chkconfig command to activate the afs configuration variable. Based on the instruction in the AFS initialization file that begins with the string #chkconfig, the command automatically creates the symbolic links that incorporate the script into the Linux startup and shutdown sequence. # /sbin/chkconfig --add afs (Optional) There are now copies of the AFS initialization file in both the /usr/vice/etc and /etc/rc.d/init.d directories, and copies of the afsd options file in both the /usr/vice/etc and /etc/sysconfig directories. If you want to avoid potential confusion by guaranteeing that the two copies of each file are always the same, create a link between them. You can always retrieve the original script or options file from the AFS CD-ROM if necessary. # cd /usr/vice/etc # rm afs.rc afs.conf # ln -s /etc/rc.d/init.d/afs afs.rc # ln -s /etc/sysconfig/afs afs.conf If a volume for housing AFS binaries for this machine's system type does not already exist, proceed to Setting Up Volumes and Loading Binaries into AFS. Otherwise, the installation is complete. afs file AFS initialization file files afs AFS initialization file Solaris AFS initialization script on client machine Reboot the machine and log in again as the local superuser root. # cd / # shutdown -i6 -g0 -y login: root Password: root_password Run the AFS initialization script. # /etc/init.d/afs start Change to the /etc/init.d directory and issue the ln -s command to create symbolic links that incorporate the AFS initialization script into the Solaris startup and shutdown sequence. # cd /etc/init.d # ln -s ../init.d/afs /etc/rc3.d/S99afs # ln -s ../init.d/afs /etc/rc0.d/K66afs (Optional) There are now copies of the AFS initialization file in both the /usr/vice/etc and /etc/init.d directories. If you want to avoid potential confusion by guaranteeing that they are always the same, create a link between them. You can always retrieve the original script from the OpenAFS Binary Distribution if necessary. # cd /usr/vice/etc # rm afs.rc # ln -s /etc/init.d/afs afs.rc If a volume for housing AFS binaries for this machine's system type does not already exist, proceed to Setting Up Volumes and Loading Binaries into AFS. Otherwise, the installation is complete. storing AFS binaries in volumes creating volume for AFS binaries volume for AFS binaries binaries storing AFS in volume usr/afsws directory directories /usr/afsws If you are using an operating system which uses packaged binaries, such as .rpms or .debs, you should allow these package management systems to maintain your AFS binaries, rather than following the instructions in this section. In this section, you link /usr/afsws on the local disk to the directory in AFS that houses AFS binaries for this system type. The conventional name for the AFS directory is /afs/cellname/sysname/usr/afsws. If this machine is an existing system type, the AFS directory presumably already exists. You can simply create a link from the local /usr/afsws directory to it. Follow the instructions in Linking /usr/afsws on an Existing System Type. If this machine is a new system type (there are no AFS machines of this type in your cell), you must first create and mount volumes to store its AFS binaries, and then create the link from /usr/afsws to the new directory. See Creating Binary Volumes for a New System Type. You can also store UNIX system binaries (the files normally stored in local disk directories such as /bin, /etc, and /lib) in volumes mounted under /afs/cellname/sysname. See Storing System Binaries in AFS . If this client machine is an existing system type, there is already a volume mounted in the AFS filespace that houses AFS client binaries for it. Create /usr/afsws on the local disk as a symbolic link to the directory /afs/cellname/@sys/usr/afsws. You can specify the actual system name instead of @sys if you wish, but the advantage of using @sys is that it remains valid if you upgrade this machine to a different system type. # ln -s /afs/cellname/@sys/usr/afsws /usr/afsws (Optional) If you believe it is helpful to your users to access the AFS documents in a certain format via a local disk directory, create /usr/afsdoc on the local disk as a symbolic link to the documentation directory in AFS (/afs/cellname/afsdoc/format_name). # ln -s /afs/cellname/afsdoc/format_name /usr/afsdoc An alternative is to create a link in each user's home directory to the /afs/cellname/afsdoc/format_name directory. If this client machine is a new system type, you must create and mount volumes for its binaries before you can link the local /usr/afsws directory to an AFS directory. To create and mount the volumes, you use the kinit command to authenticate as an administrator, followed by the aklog command to gain tokens, and then issue commands from the vos and fs command suites. However, the command binaries are not yet available on this machine (by convention, they are accessible via the /usr/afsws link that you are about to create). You have two choices: Perform all steps except the last one (Step 10) on an existing AFS machine. On a file server machine, the aklog, fs and vos binaries reside in the /usr/afs/bin directory. On client machines, the aklog and fs binaries reside in the /usr/afsws/bin directory and the vos binary in the /usr/afsws/etc directory. Depending on how your PATH environment variable is set, you possibly need to precede the command names with a pathname. If you work on another AFS machine, be sure to substitute the new system type name for the sysname argument in the following commands, not the system type of the machine on which you are issuing the commands. Copy the necessary command binaries to a temporary location on the local disk, which enables you to perform the steps on the local machine. The following procedure installs them in the /tmp directory and removes them at the end. Depending on how your PATH environment variable is set, you possibly need to precede the command names with a pathname. Perform the following steps to create a volume for housing AFS binaries. Working either on the local machine or another AFS machine, extract the Open AFS distribtion tarball onto a directory on that machine. The following instructions assume that you are using the /tmp/afsdist directory. If working on the local machine, copy the necessary binaries to a temporary location on the local disk. Substitute a different directory name for /tmp if you wish. # cd /tmp/afsdist/new_sysname/root.server/usr/afs/bin # cp -p aklog /tmp # cp -p fs /tmp # cp -p vos /tmp Authenticate as the user admin. # kinit admin Password: admin_password # aklog Issue the vos create command to create volumes for storing the AFS client binaries for this system type. The following example instruction creates volumes called sysname, sysname.usr, and sysname.usr.afsws. Refer to the OpenAFS Release Notes to learn the proper value of sysname for this system type. # vos create <machine name> <partition name> sysname # vos create <machine name> <partition name> sysname.usr # vos create <machine name> <partition name> sysname.usr.afsws Issue the fs mkmount command to mount the newly created volumes. Because the root.cell volume is replicated, you must precede the cellname part of the pathname with a period to specify the read/write mount point, as shown. Then issue the vos release command to release a new replica of the root.cell volume, and the fs checkvolumes command to force the local Cache Manager to access them. # fs mkmount -dir /afs/.cellname/sysname -vol sysname # fs mkmount -dir /afs/.cellname/sysname/usr -vol sysname.usr # fs mkmount -dir /afs/.cellname/sysname/usr/afsws -vol sysname.usr.afsws # vos release root.cell # fs checkvolumes Issue the fs setacl command to grant the l (lookup) and r (read) permissions to the system:anyuser group on each new directory's ACL. # cd /afs/.cellname/sysname # fs setacl -dir . usr usr/afsws -acl system:anyuser rl Issue the fs setquota command to set an unlimited quota on the volume mounted at the /afs/cellname/sysname/usr/afsws directory. This enables you to copy all of the appropriate files from the CD-ROM into the volume without exceeding the volume's quota. If you wish, you can set the volume's quota to a finite value after you complete the copying operation. At that point, use the vos examine command to determine how much space the volume is occupying. Then issue the fs setquota command to set a quota that is slightly larger. # fs setquota /afs/.cellname/sysname/usr/afsws 0 Copy the contents of the indicated directories from the OpenAFS binary distribution into the /afs/cellname/sysname/usr/afsws directory. # cd /afs/.cellname/sysname/usr/afsws # cp -rp /cdrom/sysname/bin . # cp -rp /cdrom/sysname/etc . # cp -rp /cdrom/sysname/include . # cp -rp /cdrom/sysname/lib . Issue the fs setacl command to set the ACL on each directory appropriately. If you wish to enable access to the software for locally authenticated users only, set the ACL on the etc, include, and lib subdirectories to grant the l and r permissions to the system:authuser group rather than the system:anyuser group. The system:anyuser group must retain the l and r permissions on the bin subdirectory to enable unauthenticated users to access the aklog binary. # cd /afs/.cellname/sysname/usr/afsws # fs setacl -dir etc include lib -acl system:authuser rl \ system:anyuser none Perform this step on the new client machine even if you have performed the previous steps on another machine. Create /usr/afsws on the local disk as a symbolic link to the directory /afs/cellname/@sys/usr/afsws. You can specify the actual system name instead of @sys if you wish, but the advantage of using @sys is that it remains valid if you upgrade this machine to a different system type. # ln -s /afs/cellname/@sys/usr/afsws /usr/afsws (Optional) To enable users to issue commands from the AFS suites (such as fs) without having to specify a pathname to their binaries, include the /usr/afsws/bin and /usr/afsws/etc directories in the PATH environment variable you define in each user's shell initialization file (such as .cshrc). (Optional) If you believe it is helpful to your users to access the AFS documents in a certain format via a local disk directory, create /usr/afsdoc on the local disk as a symbolic link to the documentation directory in AFS (/afs/cellname/afsdoc/format_name). # ln -s /afs/cellname/afsdoc/format_name /usr/afsdoc An alternative is to create a link in each user's home directory to the /afs/cellname/afsdoc/format_name directory. (Optional) If working on the local machine, remove the AFS binaries from the temporary location. They are now accessible in the /usr/afsws directory. # cd /tmp # rm klog fs vos