[Beowulf] GPFS and node ids
Many of your questions may have already been answered in earlier discussions or in the FAQ. The search results page will indicate current discussions as well as past list serves, articles, and papers.
Jose Duarte jose.duarte at human-anatomy.oxford.ac.ukFri Aug 13 07:38:27 PDT 2004
- Previous message: [Beowulf] Bioinformatics Benchmark System version 3 released
- Next message: [Beowulf] Good Grid Computing classes?
- Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]
I am just sending this slightly off topic message to the Beowulf list in the hope of anybody having come across this problem. This is a long ongoing problem that I had put aside for a while, I'm just having another go at it now. I've been trying to install IBM's GPFS parallel file system in our xServe blades (running RedHat 8). I first clone all nodes from a master image (using systemimager). Once this is done I install the rpms for GPFS and RSCT in each node separately and I compile the kernel modules successfully. But I'm hitting a problem when creating the RSCT domain as a previous step for GPFS. Apparently all nodes get assigned the same nodeid so I'm getting an error like: node33: 2632-068 This node has the same internal identifier as node30 and cannot be included in the domain definition. I suppose this is because we've cloned all the nodes. But the installation of GPFS was made from clean in each node separately and after cloning. It seems that there is some kind of "seed" from which the nodeid is created that is the same in all the clones. Also when trying the command "mknodeid" it will only say that the node has already an id assigned so I can't really change it. Anybody has seen anything similar? Any pointers much appreciated Thanks in advance Jose -- Jose Duarte MRC Functional Genetics Unit Department of Human Anatomy and Genetics University of Oxford
- Previous message: [Beowulf] Bioinformatics Benchmark System version 3 released
- Next message: [Beowulf] Good Grid Computing classes?
- Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]
More information about the Beowulf mailing list
