Anyone care to assist me with some SETI Linux set up questions?

[MereMortal]

Ok, here's the deal. I powered on our Beowulf cluster for the first time today, and noticed that the vendor had used SETI to test something (or used our cluster to increase their stats :evil: ). There is a script to send a job to the batch queue, which I'm guessing starts up 48 instances of the client.

I'm not familiar at all with running SETI, but if there isn't much reworking of the script that needs to be done, I'll let it run at least overnight to see how it fares. Now the questions:

1) Below is a segment of the script:

# Run the work unit
cd $CURDIR
./setiathome -nolock -stop_after_xfer
touch $CURDIR/stop_after_send.txt
./setiathome -nolock

while [ -f result.sah ];
do
# Failed to send the result. Keep trying.
./setiathome -nolock -stop_after_xfer
done

Will this just run one WU as it is written, or will it continue on until I stop it?

2) Can I manually set up a user_info.sah, or do I have to run the client to do that? It would be best if I could avoid running the client stopping and restarting.

 

[Confused]

You should be able to manually copy the user_info.sah file across and it will use it

As for what that script will do, I pass


Garry

 

[Smoke]

I know absolutely nothing about "Linux" but here is the "Linux Client": http://wcarchive.cdrom.com/pub/setiathome/setiathome-3.08.i686-pc-linux-gnu.tar

If this was a windows setup, all you would have to do is execute the client.

A DOS screen will open and prompt you for a response if you are a NEW user or a RETURNING user. At the following prompt, enter your email address (one that you would want to use for SETI) and let it download a WU.

This will result in the creation of many files. The one titled, "user_info.sah" is the one you will need.

I'm guessing that "$CURDIR" is the folder that the Seti Client is located in your Beowulf cluster. If it is, then you should find the other guy's "user_info.sah" file there ... just replace it.

It should then start crunching SETI for you.

I'm probably all wet with this advice but hopefully someone with Linux knowledge will chime in here. (Maybe Confused has already given you all the info you need?)

 

[BurntKooshie]

Is it still true that the Win32 client, when run under WINE, is faster than running the win32 client on a windows platform? And, is it still faster than the native linux client?

 

[Confused]

Originally posted by: BurntKooshie
Is it still true that the Win32 client, when run under WINE, is faster than running the win32 client on a windows platform? And, is it still faster than the native linux client?

From what I remember hearing (from Poof if my memory serves me correctly, I always seemed to listen when she said something ) that yes, it is faster than the native Linux client, and is almost as fast as the CLI running natively in Windows, but with VLAR WUs, it was quicker than in Windows


Confused

 

[Assimilator1]

Any luck MM?

 

[MereMortal]

I haven't had any time to devote to it today.

In the second question I was asking if one can make a user_info.sah from scratch (i.e., without ever having run the client). It's moot now, since I've run the client on a single processor and created one.

The script that was on the machine didn't seem to work, but that may be for a number of reasons. I had put a heavy firewall on the head node while I determined how I wanted to set things up. This firewall actually fubared some of the internode communication. The head node is also the only node connected to the outside world, so I might have to set up some kind proxy. I don't know enough about SETI to know if that is supported.

My boss got back today, so configuring DC is very low priority right now. I have enough work trying not to screw up the routing on this thing! The head node has four IP's and each compute node has 2 IP's. Ack, what a mess.

 

[Assimilator1]

Well SETIQ can be run as a proxy ,but you'd need to run it on windows or in WINE