Since the 15APR92 release of , the availability of data areas via NFS has vastly increased the amount of disk space accessible from a given session. The da= command line option to the aips command allows you to specify “disks” (data areas) from many hosts in addition to the current host, subject to a maximum of 35 disks per session. Note, however, that the BADDISK adverb has a limit of 10 disks. Thus, if more than 10 disks are accessed via NFS, you will not be able to prevent one or more from being used for scratch files. This can be important. Reading data over NFS is relatively efficient, but writing data is not. Even file creations (under Unix) require the writing of zeros to the whole file in order to guarantee later access to the requested space. Over NFS, this can be a slow process. For example, if user disk 1 is accessed via NFS, then every line of the message file must be written with NFS, a process which has been observed to require about one second of real time per message!
Another aspect of the new disk allocation system is a scheme by which the local Manager may restrict the availability of some disk areas to a set of user numbers, specified on a disk-by-disk basis. Managers usually use this tool to set aside most disks on a staff member’s workstation for his/her sole use and to reserve space for visitors or other special projects on “public” workstations on a case-by-case basis. Use the FREESPAC verb within AIPS to show you the space used and available on all disks for your session and also to show whether or not that space is reserved. The right-most column of FREESPAC’s output will show Alluser if the space is not reserved, Resrved if you are one of the users for which the space is reserved, Not you if you are not allowed to use the space, and Scratch if the space is to be used only for scratch files. Use FREESPAC often to keep track of how much space is available and where the space can be found.
Disk space is still generally at a premium. If more than one user has access to the disk areas you are using, then another useful tool for monitoring disks is the task called DISKU. To run it, type
This will (eventually) list on the AIPS monitor (and the message file) the amount of data space in use by each user for all disks. Identify the worst disk hogs and apply appropriate peer pressure. If you are, mysteriously, the culprit on some disk, then
to run the job, listing all catalog entries requiring more than m Megabytes.
will give you the size of your larger files on the specified disk. Armed with this information, you may be able to take appropriate action upon your own data.
Sometimes the available disk space has been eaten up by scratch files that are no longer in use. Tasks that abort while executing (and other mysterious events) may produce this situation. To delete all your scratch files, except those for tasks which are still running, type:
> SPY C R
to see which tasks are running.
> SCRD C R
to delete the files.
SCRDEST is run automatically whenever EXIT, KLEENEX, RESTART, or ABORT task_name are executed. Note that the imaging and deconvolution tasks IMAGR, APCLN, and VTESS, the data editor TVFLG, the sorter UVSRT and other tasks may create large scratch and “work” files, so you should watch for “dead” copies of scratch and work files from these programs in your disk catalog. Both MCAT and UCAT will show scratch files as well as the requested file type, but SCAT will be easier to use. Note too that, if you are using more than one computer on a given disk area, only those scratch files created by your current computer will be deleted when you run the SCRD verb. Work files have to be deleted individually since they can be still of use after the task which created them has finished.
Chapter 11 of this ookook tells you how to backup or delete your own data to relieve disk crowding. At present, all other methods for managing disk space involve system-dependent commands of one sort or another. Since these may have unexpected consequences they are not recommended.