Ok guys, it appears that it is a memory issue! I am sharing this in case anyone else encountered the same problem!
To "construct SA from BWT and Occ" is the last step in indexing. It is also the step that takes most of memory. It is possible that that node does not have enough memory and thus data keep being swapped between RAM and disk. For a 15GB reference genome, you may need around 25GB memory for this step and the subsequent mapping. If you are using LSF/SGE, please make sure you have requested enough memory.
As this is the last step, you may run: bwa bwt2sa ref.bwt ref.sa
to finish indexing, instead of running "bwa index". This step should
take several hours.
Are there any error messages? Is this the exact command you're running?
BWA doesn't pop up any error, but I am trying to find split-reads using LUMPY an it requires fasta.sa!
Are you sure you didn't mean to type
human_g1k_v37.fasta.fa
?Is reference.fa a plain multi-fasta format file? This is a straightforward command and should work.
Can run your command as and tell us what you see?
As @Devon points out below reference.fa has to be replaced with a real file name (unless that is what you file is called).
What exit code does it give?
Are you using the latest bwa?
I am using the bwa (v.0.7.12).