Transfer of updates between HEPDB and other nodes

      Updates are transferred between HEPDB and non-VM nodes using a generalisation of the above procedure, known as CDMOVE. Rather than maintain a permanent network connection with each of the remote nodes, the method is as follows:

• Upon startup, the program obtains a list of experiments for which updates are to be processed. This list is defined by the environmental variable CDGROUPS.

  Defining a list of groups

  CDGROUPS="CDCHORUS,CDCPLEAR,CDNOMAD,CDCMS,CDNA48";export CDGROUPS
  

• Each group is then processed in turn. The directory containing the NAMES file (hepdb.names) for a given group is obtained from the environmental variable CDgroup, e.g. CDCPLEAR.

• The list of remote nodes is identified by the :servers tag of the :nick.config entry. Thus, for CPLEAR, the list of remote nodes is vxcrna uxcp05 axcp01 (see below). N.B. Each database may have a different list of remote servers. The :servers tags on the individual database entries are processed by CDSERV, not CDMOVE. CDMOVE only looks at the :servers tag on the :nick.config entry.

• The remote nodes are processed for each database. Any files in the local queue for the remote node will be sent. The local queue is defined by the tag :localq. If this queue is missing, the directory tonode, e.g. tovxcrna, is used.

• The updates are transferred to the directory defined by the :queue tag. If this tag is missing, the subdirectory TODO is used.

• One may define a different list of remote nodes for each database. However, this is the concern of CDSERV, the database server, and not CDMOVE. CDSERV makes a copy of all update files in a local queue directory for each remote server defined by the :servers tag for the current database. CDMOVE looks at all of the local queues - the database to which an update file corresponds is determined by the two character prefix of the update file.

• If the tag :receive.yes is specified, any pending updates will be retrieved from the remote system.

• If the tag :poll.yes is specified, the remote node will be contacted and any pending updates will be retrieved from the remote system, regardless of whether there are any updates pending for the remote node in question.

• A connection is only made to a remote system if there are updates pending for that system. This makes sense in an environment where database updates are typically produced centrally.

• The files are received from the directory identified by the :remoteq tag. If this tag is not specified, the subdirectory queue is used.

• In all cases, the updates are received into the todo subdirectory of the local server. It is assumed that CDMOVE will be run on the same machine as the master server.

Example script to run CDMOVE

#!/bin/ksh
#
# Define the list of groups for whom updates are to be processed
#
export CDGROUPS="CDCPLEAR,CDCHORUS"
#
# For each group, define the pathname where the hepdb.names file resides
#
export CDCPLEAR=/hepdb/cdcplear
export CDCHORUS=/hepdb/cdchorus
#
# Now start the server
#
/cern/pro/bin/cdmove

NAMES file for CDCPLEAR (hepdb.names)

:nick.config
   :list.au ge ca aa
   :log./hepdb/cdcplear/log
   :queue./hepdb/cdcplear/todo
   :todo./hepdb/cdcplear/todo
   :save./hepdb/cdcplear/save
   :bad./hepdb/cdcplear/bad
   :loglevel.3
   :wakeup.60
   :servers.vxcrna uxcp05 axcp01

:nick.au
   :file./hepdb/cdcplear/aux.dbs
   :desc.auxil database
   :servers.cernvm vxcrna uxcp05

:nick.ca
   :file./hepdb/cdcplear/cal.dbs
   :desc.calibration database
   :servers.cernvm vxcrna uxcp05

:nick.ge
   :file./hepdb/cdcplear/geo.dbs
   :desc.geometry database
   :servers.cernvm vxcrna uxcp05

:nick.aa
   :file./hepdb/cdcplear/aa.dbs
   :desc.test database
   :servers.cernvm

:nick.cernvm
   :userid.cdcplear
   :node.cernvm
   :localq./hepdb/cdcplear/tovm

:nick.vxcrna
   :userid.cdcplear
   :node.axcrnb
   :localq./hepdb/cdcplear/tovxcrna
   :queue.disk$mf:[cdcplear.todo]

:nick.uxcp05
   :userid.cdcplear
   :node.uxcp05
   :localq./hepdb/cdcplear/touxcp05
   :queue.disk$mf:[cdcplear.todo]
   :poll.yes

:nick.axcp01
   :userid.cdcplear
   :node.axcp01
   :localq./hepdb/cdcplear/toaxcp01
   :queue./hepdb/cdcplear/todo

NAMES file for CDCHORUS (hepdb.names)

:nick.config
   :list.c2 ch
   :log./hepdb/cdchorus/log
   :queue./hepdb/cdchorus/todo
   :todo./hepdb/cdchorus/todo
   :save./hepdb/cdchorus/save
   :bad./hepdb/cdchorus/bad
   :loglevel.3
   :wakeup.60
   :servers.vxcrna xantia

:nick.c2
   :file./hepdb/cdchorus/charm2.dbs
   :desc.Charm2 database converted to HEPDB format
   :servers.cernvm vxcrna xantia

:nick.ch
   :file./hepdb/cdchorus/chorus.dbs
   :desc.CHORUS (geometry) database
   :servers.cernvm vxcrna xantia

:nick.cernvm
   :userid.cdhepdb
   :node.cernvm
   :localq./hepdb/cdchorus/tovm

:nick.vxcrna
   :userid.cdhepdb
   :node.vxcrna
   :localq./hepdb/cdchorus/tovxcrna

:nick.xantia
   :userid.cdhepdb
   :node.xantia.caspur.it
   :localq./hepdb/cdchorus/toxantia
   :receive:yes
   :queue./cern/hepdb/cdchorus/todo

The actual file transfer is performed by [6] routines. Thus, the remote node must be set up correctly, i.e. ZSERV must be installed. The user name and password for the remote connection are defined in the .netrc file (or ftplogin. for VMS systems) in the home directory of the account under which the CDMOVE job is run.