[Bioclusters] Sequest Cluster problems

Ken Sin Lo Ken_Sin.Lo at crchul.ulaval.ca
Fri Nov 25 15:33:34 EST 2005


We have some problems while running Sequest Cluster.  Our cluster is configured as follows:

Master Node:
  Dell Precision 670, with 2 Xeon 3GHz CPUs and 2 gb of RAM
  Microsoft Windows XP Professionnal, with Service Pack 2

Slave Node 1:
  Dell Precision 670, with 2 Xeon 3GHz CPUs and 2 gb of RAM
  Linux Fedora Core 4

Slave Node 2:
  Dell Precision 670, with 2 Xeon 3GHz CPUs and 2 gb of RAM
  Linux Fedora Core 4

We are using BioWorks version 3.1 SR1.

The first problem we have is when we start the PVM (by clicking "Start PVM" in the Cluster Manager), we always get the following error for Node 2, even if "Cleanup Nodes" has been clicked prior to starting PVM:

"Error connecting to the stderr server port!error in receive on socket: Socket shutdown by remote."

Therefore, in the PVM console, we have to type "add sequestnode2" to have Node 2 connected.  Is there any reason for this error to happen only for Node 2, knowing that Node 1 and Node 2 should be configured exactly the same way (except for the IP) ? ... Imagine if we have to manually type "add node#" 16 times, or 32 times or more!?!!

The second problem, which is unacceptable, is that when TurboSEQUEST Cluster is running, some of its slave process (sequest27_slave.exe) fails after only a few hours of computing (Windows will pop-up a "Send Report" window).  And some failures just bring down the whole search process, by freezing sequest27_master.exe.  

Does anyone have encountered the same problems and have any solution for that?  Considering the fact that we invested close to 40 000$ into this cluster (hardware and software costs), we really wish to make it works without having to restart our computers every few hours!!

And does anyone know if it's more stable to run the nodes under Windows rather than Linux?

Thank you.

Ken Sin Lo
Centre de recherche du CHUL
Quebec, Canada
G1V 4G2
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://bioinformatics.org/pipermail/bioclusters/attachments/20051125/d3461cab/attachment.html

More information about the Bioclusters mailing list