The nbi‐download partition

The nbi-download partition - retrieving external datasets

For site security reasons the HPC has very limited exposure to the external internet, which can make transferring or retrieving data onto the EI HPC storage a non-trivial problem for users. For large data transfers we always recommend using the globus service, but that is not always supported by the external body you are interacting with. For that case, the Research Computing team have configured an additional partition of nodes which all site users can access, and which has external access to the internet via the site firewall. Detailed below are two methods two use the nbi-download partition, but first please note that:

• while this service is much more open than the rest of the HPC nodes, some protocols are still unavailable (notably FTP) as a result of site security software.
• the expected use of this partition is to transfer data only, not computational workload, and because of this users are limited to 12 jobs each having 4 CPUs and 12GB of memory.
• the software node is also open to the external internet. However, this is to enable retrieving software packages not already installed in the centralised software repository and is not to be used for large downloads.

1. Interactive method to use the nbi-download partition

Often the most straightforward way to retrieve data will be using a short interactive session on the partition.

# Create an interactive session on the nbi-download partition
[aylingm@EI-HPC bash ~]$ srun -p nbi-download --mem=2GB --cpus-per-task=1 --pty /bin/bash

# load sratoolkit 2.8.2
[aylingm@EI-HPC bash ~]$ ml sra-tools/2.8.2

# Test line from https://github.com/ncbi/sra-tools/wiki/02.-Installing-SRA-Toolkit
[aylingm@EI-HPC bash ~]$ fastq-dump --stdout -X 2 SRR390728
Read 2 spots for SRR390728
Written 2 spots for SRR390728
@SRR390728.1 1 length=72
CATTCTTCACGTAGTTCTCGAGCCTTGGTTTTCAGCGATGGAGAATGACTTTGACAAGCTGAGAGAAGNTNC
+SRR390728.1 1 length=72
;;;;;;;;;;;;;;;;;;;;;;;;;;;9;;665142;;;;;;;;;;;;;;;;;;;;;;;;;;;;;96&&&&(
@SRR390728.2 2 length=72
AAGTAGGTCTCGTCTGTGTTTTCTACGAGCTTGTGTTCCAGCTGACCCACTCCCTGGGTGGGGGGACTGGGT
+SRR390728.2 2 length=72
;;;;;;;;;;;;;;;;;4;;;;3;393.1+4&&5&&;;;;;;;;;;;;;;;;;;;;;<9;<;;;;;464262

2. sbatch method to use the nbi-download partition

For large downloads, the length of time required may make an interactive session unsuitable. Instead create a job using sbatch; here I use the --wrap option, but a slurm script would function just as well.

[aylingm@EI-HPC bash ~]$ LOG_FILE=sra-test
[aylingm@EI-HPC bash ~]$ sbatch -p nbi-download -J sra-test -e ${LOG_FILE}.err -o ${LOG_FILE}.out --mem=2GB --cpus-per-task=1 --wrap="ml sra-tools/2.8.2; fastq-dump --stdout -X 2 SRR390728"
[aylingm@sub01]$ more sra-test.*
::::::::::::::
sra-test.err
::::::::::::::
Loaded module sra-tools/2.8.2
Read 2 spots for SRR390728
Written 2 spots for SRR390728
::::::::::::::
sra-test.out
::::::::::::::
@SRR390728.1 1 length=72
CATTCTTCACGTAGTTCTCGAGCCTTGGTTTTCAGCGATGGAGAATGACTTTGACAAGCTGAGAGAAGNTNC
+SRR390728.1 1 length=72
;;;;;;;;;;;;;;;;;;;;;;;;;;;9;;665142;;;;;;;;;;;;;;;;;;;;;;;;;;;;;96&&&&(
@SRR390728.2 2 length=72
AAGTAGGTCTCGTCTGTGTTTTCTACGAGCTTGTGTTCCAGCTGACCCACTCCCTGGGTGGGGGGACTGGGT
+SRR390728.2 2 length=72
;;;;;;;;;;;;;;;;;4;;;;3;393.1+4&&5&&;;;;;;;;;;;;;;;;;;;;;<9;<;;;;;464262