Harry Renshall CN/PDP
In article 1.2 of CNL 221 of December 1995 - Migration of VM User Mini-Disks and VMARCHIVE Files - we announced plans for saving both of these file bases after the closure of VM. The two filebases are being handled separately though many of the tools are common, the main component being the IBM ADSM product which is used both for the Unix public archive (pubarch) and for backing up automatically local disks of individual work stations at CERN. The VMARCHIVE occupies some 1000 magnetic tapes in the SMCG robot and these tapes are being reloaded one by one to VM disks, starting with the youngest first, then bulk copied into ADSM on a Unix server with an entry being made into a catalogue we have built. As of this date about three quarters of the tapes have been migrated. As we get to the older tapes we are seeing some that cannot be completely read and we suggest users check, as described below, that a file they believe to be vital to them has been successfully transferred. A similar procedure will be followed with the VM user minidisks and we will be starting soon to move files of users with already blocked VM accounts. Users with accounts that are still active when VM stops on the 30th June will have their files moved in the few days that follow, while we are de-commissioning the system. This will be much faster than for the VMARCHIVE as there is only one tenth as much data, about 30GB, and no tape mounts are needed. We remind you that VM Maxi-disks, which have never been backed up and which normally have addresses of 400 and above, will not be saved. Also note that only text files, ASCII or EBCDIC, will be easily understood in Unix, and users with archived compact binary PAM files in internal IBM format (CETA format is portable) or HISTORIAN or CMZ files must recall them to VM first and translate them to card image files.
The Unix tool to recall a VMARCHIVE file from ADSM is called vmretrieve and is already available for testing on those VMARCHIVE files that have already been moved. The tool relies on a catalogue of migrated files that we keep on disk for efficiency and a mapping between the original CERNVM login identifier and the afs login identifier of the same person. To use this tool you must have a CERN afs account. A user will be able to use any of his afs accounts to retrieve files from any of his CERNVM accounts, as described below, and to date about two thirds of the two thousand accounts owning VMARCHIVE files map to an afs account. It will be possible for us to add a new mapping on request even after the closure of CERNVM.
The vmretrieve command allows a user to select one of the VM userids that have been mapped to his/her afs id. In addition, a space administrator also sees VM userids belonging to the group he/she administers as an afs administrator. Only VM userids which have archived files are shown. By selecting a specific VM userid from the list, one is then shown which files had been archived, sorted by minidisks and in alphabetical order. By selecting one or multiple files, and specifying a unix name for each one, a user can then retrieve them from the ADSM database. Note that retrieve is a copy operation, i.e. the original files archived in ADSM are kept there.
At the date of writing (May 17th 1996), a beta version of vmretrieve is available on cernsp, as bjac3/public/vmretrieve. This beta version offers only a panel interface. We expect a full version to be available when this newsletter is published. This version will include a line mode and a panel interface and should be available on any PLUS service. The same functionality will be available for the VM minidisk files copied over after the closure, with a high level separation being made between the two filebases. Note that VM minidisk files of deleted accounts will not be saved as their files are deleted with the account.
Elsewhere in this CNL (section 5.9) is the announcement of pubarch, the UNIX replacement service for VMARCHIVE. We will thus be disabling the archiving of new files on CERNVM as from 1 June. Recall and query of previously archived files will still be available.
Please contact me if you have particular questions concerning your CERNVM disk files. Note that our saving of VM Minidisks is really a safety net and that users are expected to have already migrated their active working data, especially those of a binary nature, off CERNVM to either UNIX or PC environments. CNL 221 contained several articles on the procedures to be followed.