Fork() failed. Not enough space.

[dpsahoo]

I ran a parallel job with 2 nodes. "The job failed with Fork() failed. Not enough space". I checked the scratch disk for any space issues, but there was enough space.

However, when I ran the same job with 1 node it completed successfully but with a longer time.

Not sure how the change to single node config file resulted in success.

I ran the job again with 2 nodes and lo and behold it failed again with fork() failed and changing to single node file again resulted in success.

Can anybody throw light as to why a lesser config file is better than a higher node config file in certain circumstances?

[balajisr]

Check your OS Process limit configuration.

[balajisr]

Check your OS Process limit configuration.

[chulett]

With Fork() 'space' means 'memory'. You runned out.

[John Smith]

single node - your job spawn less processes hence uses less memory. configuring more nodes does not equal better unless you have resources in your box.