SLAC CPE Software Engineering Group
Stanford Linear Accelerator Center
System Admin

Managing UNCMPHST disks on VMS

SLAC Detailed
SLAC Computing
Software Home
Software Detailed
 

 

 

 


LOGICALS

File to modify: SLAC$COMMON:[MON_DISK_SPACE]MON_DISK_SPACE_START.COM

  • Logical: FREE_LIM_NOCHKLIST

"FREE_LIM_NOCHKLIST" [super,clusterwide] = "/UNCMPHST4/UNCMPHST5/UNCMPHST6/UNCMPHST7/UNCMPHST
8/UNCMPHST9/HISTRY03/HISTRY04/HISTRY08/HISTRY10/HISTRY11/SWAPNPAGE1/SWAPNPAGE2/SWAPNPAGE6/PHYSI
CSA/PHYSICSB/PHYSICSC/PHYSICSD/"

 


File to modify: SLCCOM:ASSIGNDIR.COM

  • Logical: HST_UNCMP_DATA_VMS

"HST_UNCMP_DATA_VMS" [super] = "DISK$UNCMPHST10:[PRODUCTION.HISTORY]"
= "DISK$UNCMPHST9:[PRODUCTION.HISTORY]"
= "DISK$UNCMPHST8:[PRODUCTION.HISTORY]"
= "DISK$UNCMPHST7:[PRODUCTION.HISTORY]"
= "DISK$UNCMPHST6:[PRODUCTION.HISTORY]"
= "DISK$UNCMPHST5:[PRODUCTION.HISTORY]"
= "DISK$UNCMPHST4:[PRODUCTION.HISTORY]"

 

Files to modify:

  • CLU$COMMON:[SYSMGR]DEFINE_PREFER_HSC.COM
  • CLU$COMMON:[SYSMGR]SYLOGICALS.COM
  • CLU$COMMON:[SYSMGR]SYMOUNT_DISKS.COM

 

 

General Info

Re:  Details of the NFS reorganization of "uncompressed" history files

 

  I have copied most of the VMS uncompressed history files from VMS to the NFS disk.  I

  decided to transfer files from each VMS disk to its own unique directory on the NFS disk.

  Thus, for n = 1 to 7 (and part of n = 8) I have copied

 

    DISK$UNCMPHSTn:[PRODUCTION.HISTORY]  to  DISK$NFS_SLCDATA:[PRODUCTION.HSTn]

 

  I think we should maintain this pattern in the future (for n = 8,9,10...).

 

  For history access to uncompressed files from MCC, the logicals are set up so as to get the

  more recent files from VMS, the older files from NFS.  Currently this is set up with

  n = 1,2 from NFS, n = 3,4,...8 from VMS.  However that could be changed any time to

  get up to n = 7 from NFS (and the remainder from VMS).

 

  The VMS logical names that control this file access are defined in CMS_COM_GEN

  ASSIGNDIR.COM as follows:

 

$

$         OUR_DEFINE   HST_UNCMP_DATA_VMS  disk$uncmphst8:[PRODUCTION.HISTORY],-   ! this first disk gets new files weekly!

                                           disk$uncmphst7:[PRODUCTION.HISTORY],-

                                           disk$uncmphst6:[PRODUCTION.HISTORY],-

                                           disk$uncmphst5:[PRODUCTION.HISTORY],-

                                           disk$uncmphst4:[PRODUCTION.HISTORY],-

                                           disk$uncmphst3:[PRODUCTION.HISTORY]

$!!!                                    !! disk$uncmphst2:[PRODUCTION.HISTORY],-

$!!!                                    !! disk$uncmphst1:[PRODUCTION.HISTORY]

$

$         OUR_DEFINE   HST_UNCMP_DATA_NFS  DISK$NFS_SLCDATA:[PRODUCTION.HST2],-

                                           DISK$NFS_SLCDATA:[PRODUCTION.HST1]

$!

$!  NOTE:  logical HST_UNCMP_DATA will/will-not include HST_UNCMP_DATA_NFS: when the NFS

$!         disk is/is-not available, as controlled by SLCCOM:MON_NFS_SLCDATA.SUBMIT

$!         Here, if HST_UNCMP_DATA is already defined, we leave it untouched;

$!         but if it is NOT defined (typically, at boot time), we define it WITHOUT

$!         HST_UNCMP_DATA_NFS:, and leave it up to SLCCOM:MON_NFS_SLCDATA.SUBMIT

$!         to add that if and when appropriate.

$!

$         if  f$trnlnm ("HST_UNCMP_DATA",lnmtable_name) .eqs. ""

$         then

$           OUR_DEFINE   HST_UNCMP_DATA   HST_UNCMP_DATA_VMS:

$!!!        OUR_DEFINE   HST_UNCMP_DATA   HST_UNCMP_DATA_VMS:,HST_UMCMP_DATA_NFS:  ! done in MON_NFS_SLCDATA.SUBMIT

$         endif

$

 

  Not too long from now (a month or so?) DISK$UNCMPHST8 will be full.  If you have enough

  free disk space, you could create a new disk  DISK$UNCMPHST9  and add it to the top

  of the list in HST_UNCMP_DATA_VMS in ASSIGNDIR.COM.  Alternatively, you could make sure

  the DISK$UNCMPHST1 files are backed up, delete the old files, then change its volume label from

  UNCMPHST1 to UNCMPHST9, and then use that as the new uncompressed history disk.

  (To change the volume label I think you need to DISMOUNT cluster-wide, MOUNT again privately,

  then use the SET VOLUME/LABEL= command (see HELP), dismount, then mount cluster-wide with the new label.)

 

  Note that a watchdog runs weekly (on Tues) to estimate when DISK$UNCMPHST8 will fill up.

 


 

Since we started using your new "unsparsified" history disk DISK$UNCMPHST9, I have now

copied all of the DISK$UNCMPHST8 files to the NFS disk.  Presumbably we will be doing this

again in about 8 months, and at similar intervals afterwards.

 

I've created three SLCCOM: procedures to make these future copying of files to NFS

easy to do:

 

        SUBMIT_NFS_COPY_HSTDAT.COM -- procedure to submit a batch copy job

        NFS_COPY_HSTDAT.COM        -- the batch procedure submitted by the above

        NFS_COMPARE_HSTDAT.COM     -- a procedure to verify that all files were copied

 

The batch procedures handles some details so we don't have to think about them:

        running the job under SLCSHR (which is only username with NFS WRITE priv)

        running the job on MCCDEV (which is only node that has the disk mounted for WRITE access)

        Assuring that any files previously copied are NOT re-copied

 

I've added reminders about the existence of the above procedures both in ASSIGNDIR.COM and also

in the watchdog that warns you when the current DISK$UNCMPHSTn disk is nearly full.

 



 

[SLAC CPE Software Engineering Group][ SLAC Home Page]

 

Modified: 17-Sep-2013
Created by: Ken Brobeck Sep 17, 2013