Trinity error around number of CPU's and Java virtual machine

[GenoMax]

Also, what exactly is Jellyfish? Is it a specialized program for counting k-mers in Trinity? Or is it used outside of Trinity? Does --JM refer to the overall memory allocation, or to the memory allocation per thread?

Nature protocols article: http://www.ncbi.nlm.nih.gov/pmc/articles/PMC3875132/

Stages within Trinity (http://trinityrnaseq.sourceforge.net...ity_guide.html)

Stage 1: generate the kmer-catalog using jellyfish
Stage 2: run Inchworm to generate draft contigs
Stage 2: Chrysalis clustering of inchworm contigs and mapping reads
Stage 3: Chrysalis deBruijn graph construction
Stage 4: Run butterfly, generate final Trinity.fasta file

[gwilymh]

Thank you all for the responses. Where can one find the documentation for the butterfly-specific (--bfly*) options? There does not seem to be any in the Trinity manual, webpage or --help instructions.

Also, what exactly is Jellyfish? Is it a specialized program for counting k-mers in Trinity? Or is it used outside of Trinity? Does --JM refer to the overall memory allocation, or to the memory allocation per thread?

[GenoMax]

8 cores and 16G RAM is not a lot, specially if there are other jobs running on that node. Are you using a queuing system to submit the jobs?

[westerman]

Has anyone else had trouble with executing Trinity over more than two CPUs?

No. People run Trinity with more than two CPUs all of the time.

Since you have your cluster support people working with you they would be the best people to troubleshoot your problem since I suspect a cluster setup problem. My first suspicion -- and this is true anytime someone says they are having problems using more the 4GB -- is to look for a 32-bit dependency somewhere. Specifically what version of Java are you using?

BTW: '-JM' only applies to Jellyfish memory and has nothing to do with Java. Look at the '--bfly*' options for some Java parameters.

[gwilymh]

Hi,

I have been sending trinity jobs to a cluster under differing conditions. If I request 1 node with one or two processes (CPU's) and 2GB of memory for each process/CPU, the program runs without a hitch.

However, If I request four processes/CPU's from a single node or from two nodes, the program fails with the following error message:

initialization of VM
Error occurred during initialization of VM
Could not reserve enough space for object heapCould not reserve enough space for object heap

Could not create the Java virtual machine.
Could not create the Java virtual machine.

In all runs, I match the --JM flag and --CPU flag in Trinity with the number of processes and the amount of memory requested from the cluster.

Each node in the cluster has 8 CPU's and a total of 16GB of memory, so I am by no means overburdening the system. The cluster support team tested this by running Trinity input with a total of 24G of memory and got the same answer.

Has anyone else had trouble with executing Trinity over more than two CPUs?