Running AIPS at the ATNF

This page documents how to run AIPS at ATNF's sites. It only covers getting the software running. If you are having troubles with setting AIPS task parameters etc, you need to look elsewhere.

[ Availability | Starting | Linux/Solaris | Troubleshooting ]

Availability

The new version of AIPS should be available on most Solaris and Linux hosts. Some machines may need a small amount of setup before they can run AIPS. Please write to aipsmngr_AT_atnf.csiro.au for assistance with this. It won't take long to do.

Please note: support for Solaris is being phased out.

Suggested hosts

The following server machines are good choices for running AIPS. To make use of them you will need to read about how TV servers work.

Hostname	OS	Site
draco.atnf.CSIRO.AU	Linux	Marsfield
hydra.atnf.CSIRO.AU	Linux	Marsfield
serpens.atnf.CSIRO.AU	Linux	Marsfield
kaputar.atnf.CSIRO.AU	Linux	Narrabri
pices.atnf.CSIRO.AU	Solaris	Parkes

Starting AIPS

You should be able to start aips by just typing

  unix% aips

If you haven't run AIPS before you should also read about TV servers.

ATNF-Specific AIPS Tasks

The ATNF specific task ATLOD and ATNF procedures (ATCALIB etc) will not work the first time you start AIPS. You will need run the following two commands after starting up:

   > run newparms
   > run atprocs

After that, you should be able to use the AT-specific tasks.

Sharing between Linux and Solaris Disks

Please note: support for Solaris is being phased out.

AIPS is able to access NFS mounted disks from either Linux or Solaris hosts. However data and catalog files are sensitive to the byte ordering used by the CPU that AIPS is running on. This difference in interpretation is often called "endian-ness".

This means that you should take care to access a given /DATA/HOST_N/AIPS storage area from either Solaris or Linux, but not from both. It may be possible to corrupt the catalog files if you get this wrong.

If you get lots of horrible messages regarding MSG files etc when you start AIPS, you are trying to access your data the wrong way around. This should not hurt anything if you exit AIPS at once, but it may. So our advice is to exit AIPS and check which OS is running on the machine hosting the disk you want to access. To check the OS that is running, use the command uname -a.

We have tried to address this by keeping two separate AIPS "sites" for the new version: "ATSOL" for Solaris, and "ATLIN" for Linux. This means that the data disks you can see from AIPS should be limited to those of the same type, which can be used safely together. So you shouldn't see any warnings such as those noted above. If you do, please contact aipsmngr.

The two sites have a different "master" host - cetus for Solaris, and hydra for Linux. This means your task parameters will be saved separately between Solaris and Linux. For example if you try running IMAGR on a Solaris machine, and then try to TGET the same parameters on a Linux machine, you won't get the ones you just set, but rather the ones you used last time on Linux.

Troubleshooting

I keep getting password prompts

AIPS now relies heavily on SSH for sending data between hosts. This has some advantages but requires a little setup on the part of the user.

Aips freezes at startup

AIPS may freeze after printing a message like this:

   START_AIPS: Checking if vulcan.atnf.CSIRO.AU is reachable via ssh...

This happens because AIPS is trying to use SSH to connect from the remote machine to back your local desktop. A freeze at this point indicates the SSH connection is is not working.

Try manually making a SSH connection from the remote server back to your local workstation. If that works, re-read the documentation on TV Servers. If manual SSH connection fails, seek help from ServiceDesk.

The TVSERVER does not start (Solaris)

Look for the line about shared memory in the startup messages. This means that aips could not allocate enough shared memory for the TV server. There are two solutions:

Short term: Turn off shared memory for the TV server. This slows things down a bit, but not much.
```
      local> xrdb -merge /nfs/AIPS/xas.settings.
    
```
That file only has one setting in it - you could add it to your own .Xdefaults if you wish.
```
       AIPStv*useSharedMemory: 0
    
```
This only needs to be run once after restarting your X session. Running it more than once won't hurt, but won't help either.
Fix the problem. Send an email to ServiceDesk requesting that the shared memory chunk size be increased. The machine will then have to be rebooted. You should request that the following lines be added to /etc/system
```
    set shmsys:shminfo_shmmax=8388608

    set ufs:ufs_HW=6291456
    set ufs:ufs_LW=4194304
    set priority_paging=1
    
```
The last three lines are optional, but they improve speed generally for AIPS.

I get errors such as "Xlib: connection to "vulcan.atnf.CSIRO.AU:0.0" refused by server"

First, check you haven't exceeded your quota in your home directory. X uses a file called ~/.Xauthority to authorise connections to X servers. If you are over quota, X cannot add another entry to this file, and you will get messages like the one above. A quick way to fix this is to delete one file in your home directory.

Otherwise, on your local machine try:

     local> xhost +<localhostname>

If that fails try:

     local> xhost +<remoteservername>

ssh-agent is set up OK but AIPS still stops connecting to the local machine.

This happens rarely. It occurs because AIPS uses 'short' machine names instead of fully qualified domain names (FQDN). To check if this is the problem, try making a SSH manually from the remote machine to the local machine.

If that works, most likely one or the other ends is searching the wrong DNS domain for the short name. Contact ServiceDesk and ask them to check the default DNS search domain for the remote machine and your local machine.