Hi.
I can get megahit (v1.1) to work as expected (either by compiling it myself or using precompiled binaries) on our cluster with CentOS (6.8).
The same version of megahit initially appears to run on Ubuntu (14.04) but no contigs are assembled and there are no errors displayed. The command I am using for testing purposes is:
On Ubuntu, the screen only displays:
..and the resulting assembly folder is empty.
If I execute the same command line instructions as above on Centos (on our cluster) the screen displays:
...and lots more output as the program correctly executes and assembles the contigs.
I'm guessing that perhaps megahit is coded to abort if there is not a minimum amount of memory available on the machine??
Usually I like to construct pipelines etc on my small virtual machine before porting them over to the cluster, hence my attempts to get them to work on this low memory machine.
Have I come to the correct conclusion or is there some other possibility (and associated fix) that exists?
I can get megahit (v1.1) to work as expected (either by compiling it myself or using precompiled binaries) on our cluster with CentOS (6.8).
The same version of megahit initially appears to run on Ubuntu (14.04) but no contigs are assembled and there are no errors displayed. The command I am using for testing purposes is:
Code:
megahit -t 1 -m 1000000000 --verbose -r /home/input/readsInterleaved3.fa -o /home/output/assembly
Code:
MEGAHIT v1.1.1 --- [Wed Jan 18 15:32:08 2017] Start assembly. Number of CPU threads 1 --- --- [Wed Jan 18 15:32:08 2017] Available memory: 2099187712, used: 1000000000
If I execute the same command line instructions as above on Centos (on our cluster) the screen displays:
Code:
MEGAHIT v1.1.1 --- [Wed Jan 18 15:45:15 2017] Start assembly. Number of CPU threads 1 --- --- [Wed Jan 18 15:45:15 2017] Available memory: 126861316096, used: 1000000000 --- [Wed Jan 18 15:45:15 2017] Converting reads to binaries ---
I'm guessing that perhaps megahit is coded to abort if there is not a minimum amount of memory available on the machine??
Usually I like to construct pipelines etc on my small virtual machine before porting them over to the cluster, hence my attempts to get them to work on this low memory machine.
Have I come to the correct conclusion or is there some other possibility (and associated fix) that exists?