CERN-SL ACCELERATOR PHYSICS GROUP


SLAP Work Group Service


History

For several years the SL/AP group had its own computing facilities. The machines (3 HP9000/730/735 UNIX servers and a small number of HP9000/710/720 desktop systems) had a customized setup in order to meet the special requirements of the users. This included:
Root access to all machines in order to perform customizations and system control (M. Böge, H. Grote and C. Iselin)
Customized shell setup in order to provide certain environment settings like default printers, default compiler switches, default path settings etc.
A customized HP VUE based desktop with the the HP VUE file manager `vuefile'
A network file system (NFS) which allowed access to user data physically located on one machine from all other machines
A central NIS user registry
Central installation of programs which were used by a large fraction of the group like MAD. At that time the anonymous account "slath" was used for this purpose
The /usr/local tree was provided by the central ASIS service
"Batch service" on a "good will" basis using proper nice values for long running processes
A WWW service based on the CERN http daemon (Home URL: http://hpariel.cern.ch/) for group related information and documentation
Boot support for X terminals owned by the group
Font service for the X terminals
Easy access to certain hardware components like cdrom drives, magneto optical drives and tape drives
Over the last years the central computing support has improved a lot and solutions were found which allow the group specific customization of central services (Work Group Server (WGS) concept and the HEPiX environment). At some point the AP group decided to give up on its own servers and to go for a WGS solution (SLAPWGS) which is managed by people from the IT division (SLAPWGS: Tim Smith). Although the management is central the use of the service is restricted to members of the AP group. Accounts are available on request (H. Grote).

The actual service consists of 2 HP D-Class SMP Servers with 2 160 Mhz PA-8000 64-bit processors with 512 Mb physical memory and 8Gb disk space. The machines involved are SLAP01 and SLAP02. A third machine has been ordered recently which will operate as a batch server where the other machines are for interactive use ONLY. The dispatcher SLAP.WGS.CERN.CH alias SLAP forwards users to the least loaded machine.


SLAPWGS

Owing to the fact that the list of requirements has not changed the customization of SLAPWGS ( ) reflects somehow the customization which was done for the old HP service ( ). In the following a comparison of the old ( ) and the new service ( ) is performed :
Root access to all machines in order to perform customizations and system control
On SLAPWGS only selected people from the IT division have root access (Tim Smith, Tony Cass). Members of the local group "slapstaff" (slapstaff::82:fci,hansg,ghislain,boege,jsberg) are allowed to execute the script `killsi' invoked using `sudo' in order to kill processes of users belonging to the group "si". All customizations are done on the Group and Cluster level of the HEPiX environment. This reduces the amount of system administration work to a minimum and provides a homogenous machine setup
Customized shell setup in order to provide certain environment settings like default printers, default compiler switches, default path settings etc.
The Group level setup (see (5) in figure 1) foreseen in the HEPiX environment provides the members of the "si" group with a customized environment on all HEPiX conform nodes. The Cluster level setup (see (4) in figure 1) allows a customized setup of the Work Group Service SLAPWGS (See Group level setup and Cluster level setup)
A customized HP VUE based desktop with the the HP VUE file manager `vuefile'
SLAPWGS confronts the user with a similarly customized `fvwm' based desktop. The default File Manager is `moxfm' (See Desktop setup)
A network file system which allows access to user data physically located on one machine from all other machines
On SLAPWGS the standard network file system is AFS. All user data reside on central AFS volume servers (See AFS related information). The only exception is the /scratch/si area which is meant for storage of temporary data. The NFS file system of the old HP service is available for a short transition period (See NFS related information)
A central NIS user registry
On SLAPWGS central CERN user registration is performed via `(x)userreg'. User passwords are AFS passwords. Owing to the fact that the old HP service is still alive an automatic user registration procedure has been developed. Newly registered users on SLAPWGS get their account on the old HP service about one hour after their SLAPWGS registration
Central installation of programs used by a large fraction of the group like MAD
All group specific data are kept with respect to the HEPiX supported "si" group directory /afs/cern.ch/group/si (See Group directory structure)
The /usr/local tree
On SLAPWGS the /usr/local tree is provided by the central ASIS service. To gain performance some of the frequently invoked packages are kept locally with respect to /asis.local (for example the emacs package)
Batch service
There is no "batch service" on a "good will" basis on SLAPWGS. Long running processes are reniced and finally killed by the WGSMON service. It is possible to launch batch jobs on CERNSP and SHIFT54 remotely from SLAPWGS (See Batch service)
WWW service for group related information and documentation
SLAP01 runs a WWW service on port 80 (Home URL: http://wwwslap.cern.ch/). The service is based on the APACHE http daemon (See WWW service)
Boot support for X terminals owned by the group
SLAP01 and SLAP02 provide a boot service for HP and NCD X terminals via TFTP (See X terminal support)
Font service for the X terminals
SLAP01 runs the font server `fs'on port 7100 (See Font service)
Easy access to certain hardware components like CDROM drives, magneto optical drives and tape drives
On SLAPWGS it mainly concerns the access to a magneto optical drive which is physically connected to a machine named HPSLAP03 which is the surviving of the old 3 HP9000/730/735 servers (See Hardware access)

Problems?

if you have problems try to solve them by consulting the following documentation. In case of severe problems you may contact the system administrator of SLAPWGS Tim Smith.
The nice Micro Guide for HP Work Stations in the SL/AP Group by H. Grote and C. Iselin is still available (-> postscript version)
Have a look at:
this document ;-) (-> postscript version)
the introduction "UNIXhelp for Users". You can search the document if you have specific problems
the small collection of links on this WWW server
the CERN UNIX User Guide
John's collection of UNIX links
what the DESY people provide (USG Documentation)


MB

Group level setup

There are four customized HEPiX scripts on the Group level (see (5) in figure 1 and figure 1a) which are located with respect to /afs/cern.ch/group/si/slap/etc/hepix. They are linked back to /afs/cern.ch/group/si which is the standard HEPiX location for these files:
group_sys.conf.sh and group_sys.conf.csh: These startup scripts for Bourne (.sh) and C (.csh) shells define some important environment variables for all shells (interactive and noninteractive):

AP_GROUP_CONF=true Group environment loaded !
AP_GROUP=slap Identifier of AP group is "slap"
AP_GROUP_DIR=$GROUP_DIR/$AP_GROUP (HEPiX: GROUP_DIR=/afs/cern.ch/group/si) The AP group home
AP_PTS_SLAP=true The invoking user is member of the protection group slath:slap
AP_WWW_HOME=http://wwwslap.cern.ch/ The home URL of the WWW server
AP_WWW_HOME_DIR=$AP_GROUP_DIR/etc/httpd The ServerRoot of the WWW server
AP_HARVEST_HOME=$AP_GROUP_DIR/etc/etc/harvest The home directory of the Harvest Broker

LHC_DIR=$AP_GROUP_DIR/project/lhc The LHC project home
GROUPPATH=$AP_GROUP_DIR/bin:$AP_GROUP_DIR/scripts:$GROUP_DIR/bin:$GROUP_DIR/scripts The additional group path which gets merged into PATH
MANPATH=$AP_GROUP_DIR/man:$GROUP_DIR/man:$MANPATH The manual path extension
XFILESEARCHPATH=$AP_GROUP_DIR/lib/X11/app-defaults/\%N:$GROUP_DIR/lib/X11/app-defaults/\%N:\
/usr/lib/X11/app-defaults/\%N:/usr/local/lib/X11/app-defaults/\%N The search path for X application defaults

TEXMFS="{$AP_GROUP_DIR/share/texmf,"'!!$TEXMF}' add the location of group specific TeX style and class files to the "kpathsea" search path
TEXINPUTS=.:$HOME/share/tex/inputs//: add the location of user specific TeX style and class files to the "kpathsea" search path
group_login.sh and group_login.csh: These startup scripts for Bourne (.sh) and C (.csh) define things to be done in login shells:

AP_GROUP_LOGIN=true Group login passed !
/usr/local/bin/qset -h shift54 Define SHIFT54 as the default NQS remote host

figure 1: Steps of the HEPiX login

Cluster level setup

There are 12 customized HEPiX scripts on the Cluster level (see (4) in figure 1) which are located with respect to $AP_GROUP_DIR/etc/hepix/cluster. On SLAPWGS the standard HEPiX location /etc/hepix/cluster is a link to this directory:
sys.conf.sh and sys.conf.csh: These startup scripts for Bourne (.sh) and C (.csh) shells define some important environment variables for all shells (interactive and noninteractive):

AP_CLUSTER_CONF=true Cluster environment loaded !
GROUP_DIR=/afs/cern.ch/group/si Force AP group environment for non "si" users ("$GROUP" != "si")
PRINTER=30_5_cor_hp_ps Define default printer
AP_CLUSTER_TMP=/scratch/si Location of local scratch space
env.sh and env.csh: These startup scripts for Bourne (.sh) and C (.csh) shells define some important environment variables for all shells (interactive and noninteractive):
add contents of $AP_GROUP_DIR/etc/hepix/cluster/PATH to PATH and $AP_GROUP_DIR/etc/hepix/cluster/MANPATH to MANPATH
login.sh and login.csh: These startup scripts for Bourne (.sh) and C (.csh) define things to be done in login shells:
cat $AP_GROUP_DIR/etc/hepix/cluster/motd Show motd
$AP_GROUP_DIR/bin/news-slap -n show SLAP news items
zshrc and tcshrc:
Define some aliases and functions for `zsh' and `tcsh'
xdm/Xresources:
Set greeting resource to SLAP WGS
xprofile:
HX_WM="/afs/cern.ch/group/si/slap/bin/fvwm-slap" Default window manager on SLAPWGS is `fvwm-slap' (fvwm version 1.24r)
HX_LAST_CLIENT="/bin/sh -cf /afs/cern.ch/group/si/slap/lib/session/session" Default last client is SL/AP session manager `session'
xclients:
Load Xresources by sourcing $AP_GROUP_DIR/scripts/x11init (Note: we use $AP_GROUP_DIR/etc/hepix/cluster/.Xresources instead of the HEPiX conform $AP_GROUP_DIR/etc/hepix/cluster/Xresources)
The load hierarchy of Xresources redefined by `$AP_GROUP_DIR/scripts/x11init' looks as follows:
/usr/local/lib/hepix/X11/hep/HEP_Xresources HEP level
/usr/local/lib/hepix/X11/site/Xresources Site level
/etc/hepix/cluster/Xresources Cluster level (NOT used)
$AP_GROUP_DIR/etc/hepix/cluster/.Xresources Cluster level
/etc/hepix/Xresources System level (NOT used)
$HX_GROUP_DIR/group_Xresources Group level (NOT used)
~/.hepix/xresources User level
~/.Xresources User level
Start a few X clients ( `xterm', `datebook', `xcalc' and `xclipboard' )
xclients.m:
Perform xhost +hostname for our HP's used as X terminals
Load keyboard map $AP_GROUP_DIR/etc/hepix/cluster/Xmodmap.HPP for our HP's used as X terminals
Add default fontserver tcp/xtsoft2:7100 if not already in font path
Add default fontserver tcp/slap01:7100 if not already in font path

figure 1a: Startup files of the HEPiX login


Desktop setup

The "feable" window manager `fvwm'
Fvwm is a "light weight" window manager with "virtual" window features via the FvwmPager module.
The configuration file system.fvwmrc and the modules are located with respect to $AP_GROUP_DIR/lib/X11/fvwm
In order to customize fvwm copy $AP_GROUP_DIR/lib/X11/fvwm/system.fvwmrc to ~/.fvwmrc.
Please consult the manual pages for fvwm, FvwmPager, GoodStuff, FvwmBacker and FvwmScroll for details on the format of the configuration file
The file manager `moxfm'
Moxfm is an extended Motif version of the file manager `xfm' with features very similar to those of the HP VUE file manager `vuefile'
The configuration files are located with respect to $AP_GROUP_DIR/lib/X11/moxfm/dot.xfm. All files are automatically copied to ~/.xfm on the first invocation of the program. Refer to the moxfm manual page to learn about the syntax of the configuration files

AFS related information

To get used to the AFS terminology refer to the CERN AFS User's Guide and the AFS Disk Space Administration guide
To speed up AFS performance by increasing the cache cycle time on SLAPWGS the AFS cache size was set to 200000 1K byte blocks instead of 82365 1K byte blocks on CERNSP
There are three group related AFS protection groups under the control of the anonymous user "slath":
slath:slap All AP group members are members of this group
Members have to be added/removed using the scripts `adduser_slap'/`removeuser_slap' (See also AP_PTS_SLAP)
slath:slapstaff The members of this protection group have write permission with respect to $AP_GROUP_DIR. Be careful adding people to this group !
slath:wwwslapstaff The members of this protection group have write permission with respect to the ServerRoot directory $AP_WWW_HOME_DIR
The concept of having a ~/public and ~/private directory has been extended to the ~/group directory which gives all AP group members read access to files with respect to this directory. All group members should create ~/group and give the correct permissions using the command `allow_slap ~/group'. Furthermore members are encouraged to create lep and lhc directories with respect to ~/group !
Locking the screen using the SL/AP session manager `session' leads to AFS token refresh ! Locking and unlocking the screen within less than 25 hours which is the expiration time of the token guaranties a valid token !

NFS related information

The user data of the old HP service can be found with respect to /nfs/ followed by the node name. Up to now seven nodes survived: hpanne, hplear, hpslap01, hpiago, hppuck, hpslap03, romeo
Recently a 10 processor Alpha Server (node name: SHIFT54) was acquired which has NFS based home directories. These home directories can be accessed from SLAPWGS via /u/si/ followed by the username. The scratch space /scratch on SHIFT54 can be accessed via /nfs/shift54/scratch (See figure 2)

Group directory structure

Owing to the fact that "si" covers more or less the entire SL division the AP group information is not kept with respect to the HEPiX "si" Group home /afs/cern.ch/group/si but with respect to $AP_GROUP_DIR which is /afs/cern.ch/group/si/slap (see Group level setup)
The directory structure with respect to $AP_GROUP_DIR looks as follows:

bin          Binaries
doc          Documents
etc          WWW service and HEPiX stuff
home         Links to members of the protection group slath:slap
include      Include files
info         Emacs info pages
lib          Platform dependent data, libraries
man          Manual pages
project      Pointers to group related projects
scripts      Platform independent scripts 
share        Platform independent data
src	     Source files
tmp	     Temporary files
The old "slath" account is still available but reserved for the LEP optics database under ~slath/machines

Batch service

LoadLeveler service
The LoadLeveler queue on CERNSP is available from SLAPWGS using the usual LoadLeveler commands `llq', `llsubmit' and `llcancel' (See figure 2)
On SLAPWGS the LoadLeveler clients are pointers to the wrapper script `remsh_cernsp' which executes the "real" clients remotely on CERNSP
For the submission of MAD jobs the script `madbatch.ll' can be used
For the submission of SIXTRACK jobs the graphical interface `6T' invoked with the argument ll can be used
NQS++ service
The NQS++ queue of SHIFT54 is available from SLAPWGS using the usual NQS commands `qstat', `qsub' and `qdel'
On SLAPWGS the NQS clients are pointers to a wrapper script which invokes `qtell' to execute the "real" clients on the default NQS remote host SHIFT54
For the submission of MAD jobs the script `madbatch.nqs' can be used
For the submission of SIXTRACK jobs the graphical interface `6T' invoked with the argument nqs can be used

figure 2: SLAPWGS and other services


WWW service

To get an overview have a look at the SLAP WWW Service page
The binaries and configuration files are located with respect to the "ServerRoot" directory $AP_WWW_HOME_DIR. On SLAP01 this location is also reachable via /usr/local/etc/httpd
The "DocumentRoot" is $AP_WWW_HOME_DIR/htdocs which corresponds to the Home URL http://wwwslap.cern.ch/.
The server process (-> status) is controlled with the script `HTTPd start|stop' invoked using `sudo'. The members of the group "slapstaff" (slapstaff::82:fci,hansg,ghislain,boege,jsberg) are allowed to take this action
In order to help people to migrate to the new service a "fake" server has been introduced on HPSLAP03 redirecting people from http://hpariel.cern.ch/ to http://wwwslap.cern.ch/
A couple of scripts is periodically invoked under the user "boege" on SLAP01 in order to do some statistics on the server load and to retrieve publication lists from the preprints server (See $AP_WWW_HOME_DIR/cron/crontab.slap01 for details)
The directory $AP_WWW_HOME_DIR/Incoming which is local to SLAP01 contains temporary data like web counter files and the SL/AP calendar database
The Apache URL Rewriting Module is used to fake an "xwho" service on this server (the "xwho" service is used on the SL page AP Group). New group members have to be added to the file $AP_WWW_HOME_DIR/htdocs/who/people/.htaccess !
To learn more about the server configuration refer to the Apache documentation and the Apache URL Rewriting Module document

The Harvest Broker software with Glimpse as indexing interface is running on SLAP01 on port 8500 and 8501 (-> Server Summary). The broker provides searchable information about web pages which are available via wwwslap.cern.ch. Specific search keys make it possible to restrict the search (examples are the MAD8 and SIXTRACK documents)
The binaries and configuration files are located with respect to the "HARVEST_HOME" directory $AP_HARVEST_HOME. On SLAP01 this location is also reachable via /usr/local/etc/harvest
broker, glimpseserver and gatherd are running under the anonymous user "harvest" (See ~harvest/cron/crontab.slap01 for details). The "gatherer" database is updated every week. Queries to the database can be performed using the form http://wwwslap.cern.ch/search/

X terminal support

The server software and the configuration files are located with respect to /usr/tftpdir
All X terminals should have "SLAP01" and "SLAP02" as primary and secondary boot servers with boot method "TFTP" selected. There is no support for anonymous NFS (See X terminal guide for details)
Most of the old HP9000/710 machines have been reconfigured as X terminals. They run the simple chooser script xtrc from /etc/inittab. There is also support for an X based chooser (Get chooser-HPUX.tar)

Font service

The configuration file config is located with respect to $AP_GROUP_DIR/lib/X11/fs
The fonts reside relativ to the fonts directory $AP_GROUP_DIR/lib/X11/fonts
The font server is running under the anonymous user "harvest" on SLAP01 (See ~harvest/cron/crontab.slap01 for details)

Hardware access

Magneto optical disk
There are two scripts which allow the optical disk to be mounted from SLAPWGS namely `mount_od' and `umount_od'. The mount point on HPSLAP03 is /MO. This filesystem is exported to SLAPWGS and NFS mounted under /OPDISK. Users can access their data via this NFS mount. Both scripts are invoked using `sudo' because they need to be executed as "root". Thus the invocation requires the specification of the AFS password


MB
Last Updated November 30, 1997

Back

Back to Computing