Laserfiche WebLink
Another way to implement NFS is to export the top-level directory associated with disk partitions <br />and then use symbolic links on each machine to map common directories to the machine-dependent <br />directories. When using this method, it is important that disk partitions be uniquely named for each <br />machine. For example, rather than always using , , etc., for disk partitions, use the <br />/usr2 /usr3 <br />machine name in disk partition names, e.g., , ("proto" is the name of the <br />/proto1 /proto2 <br />prototype machine at RTi that is used to preview the prototype system to be installed on the Briefing <br />Room machine). This requires that machine names generally do not end in a number. Each machine <br />that accesses disks from "proto" using NFS has a directory called which contains mount points <br />/nfs <br />for the "proto" disks: <br />/nfs/proto1/ <br />proto2/ <br />This directory structure is the same on every machine in the network. Because entire partitions from <br />"proto" are exported, every file in those partitions is available. However, in order to make disk <br />access more convenient, symbolic links are used to create common directories (see Task <br />Memorandum 1.05-23 for a summary of the CRDSS file structure.) For example, if all DMI source <br />code is on machine "nave" in directory /nave1/libdmi, then "proto" would have the following NFS- <br />related directory structure: <br />/nfs/nave1/ <br />nave2/ <br />/crdss/libdmi->/nfs/nave1/libdmi <br />The -> notation in UNIX indicates that the directory (or file) to the left of the arrow actually exists as <br />the directory (or file) to the right of the arrow. Users who change to the directory /crdss/libdmi on <br />"proto" then only have to change to that directory. The use of the directory on "nave" will not be <br />evident. <br />If this standard method for exporting partitions from a machine is used, then there are no major <br />limitations to exporting every partition on every machine (except the root partition). Where <br />machines have common partition names (e.g., ), the directory must be referred to differently on <br />/usr <br />remote machines (e.g., ). Mounting all remote disks on a single machine will be <br />/nfs/nave-usr <br />used on the machine that performs backups so that backups can be done as if every file is on a single <br />machine. <br />The CRDSS files will be divided among machines according to major functional requirements, and <br />limited by the number of machines that are available. For example, the database server machine will <br />primarily contain the INFORMIX database and raw files that are to be loaded in the database. Other <br />machines will contain source code under development. The file structure described in Task <br />Memorandum 1.05-23 will be implementing by using NFS and symbolic links to distribute files <br />among the CRDSS machines. <br />All user accounts on the CRDSS workstations at RTi will be located on a single machine (or split up <br />logically). NFS and symbolic links will be used to make the user accounts available from each <br />CRDSS machine. Network Information System programs will be used to distribute password files <br />between the machines. <br />2 <br />A275 05.10.94 1.05-25 Malers <br />