velvetg problem after Threading through reads

Question

[avolkov]

Hi all!

I am de novo assembling a plant transcriptome from 430M quality-trimmed 2x100 Illumina reads on a 24 CPU 156 GB RAM machine using velvet. Everything was fine until this velvetg step:

.....
[16974.644928] Threading through reads 179000000 / 430800790 
[17586.290922]  === Threaded in 9824.522651 s

Here I expect the assembly o proceed to the next step, but it does not. I waited for many hours but nothing happens. The process is still running, but does not use any CPU. However, memory is 100% in use and 48G in swap. Disk is at 250 k/s reading from swap.

What should I do? Shall I stop the assembly? Computer time is expensive so I don't want to wait for something that will not happen.

Thanks in advance!

Answer 1

[Antonio R. Franco]

Iy seems that you run out of memory ant the program is using a virtual disk. The whole population of der Bruijn nodes (kmers) are usually kept in memory

If so, you need to be alerted that memory works a lot much faster (in the order of nanoseconds) than when used a disk (sometimes in the order of milliseconds).

If you are aware of a very high activity of the disk, this should be the reason of your delay

[avolkov]

Thanks for the answer. The problem is that there is barely any disk activity. But i also believe that this problem is memory-related. Still, it looks like the software is not doing anything, since the message about the next step is usually there instantly after threading. For now i stopped the assembly. Will have to retry with more memory and perhaps a fast ssd disk.