Find a way to override Spark memory allocation from docker run command

[davidonlaptop]

Memory allocation is hardcoded in adam-submit launch script

If there is not enough memory allocated to Adam / Spark, Adam may crash. By default, only 512MB is allocated to java processes. Here's a workaround to fix this issue

Download a larger BAM file and start a bash shell in an ADAM container:

mkdir -p ~/data/1kg/samples/hg00096/
wget -nH --cut-dirs=99 \
     -P ~/data/1kg/samples/hg00096/  \
     ftp://ftp.1000genomes.ebi.ac.uk/vol1/ftp/phase3/data/HG00096/alignment/HG00096.chrom20.ILLUMINA.bwa.GBR.low_coverage.20120522.bam

Then, start an ADAM container:

docker run --rm -ti -v $HOME/data/1kg/samples/hg00096:/data gelog/adam bash
root@42c257dcfbcc:/# 

Once inside the container, run ADAM with 1.5GB of RAM (the SPARK_DRIVER_MEMORY and SPARK_EXECUTOR_MEMORY variables):

root@42c257dcfbcc:/# SPARK_DRIVER_MEMORY=1500m SPARK_EXECUTOR_MEMORY=1500m \
    adam-submit transform \
    /data/HG00096.chrom20.ILLUMINA.bwa.GBR.low_coverage.20120522.bam \
    /data/HG00096.chrom20.ILLUMINA.bwa.GBR.low_coverage.20120522.adam
root@42c257dcfbcc:/#

[UPDATE: June 2nd 2015]: The adam-submit script in Adam 0.16.0 now allocates 4 GB by default see here. If you need more than 4GB, then we need to find a way to override this parameter when running the container. The adam-submit script accepts the --conf spark.executor.memory="1g" option, but it has no effect.

Please contact us if you find a more elegant way to solve this issue.

[akhmees]

I am not sure if this is correct. I am started adam-submit on docker without specifying any parameters and it only allocated 256MB of memory.
#adam-submit -- flagstat NA12878.adam 16/09/28 06:15:17 INFO MemoryStore: MemoryStore started with capacity **265.4 MB** 16/09/28 06:15:19 INFO MemoryStore: ensureFreeSpace(171744) called with curMem=0, maxMem=**278302556**
I can change the memory by adding spark args but surprisingly i don't see any performance gain. ADAMN=2.10-0.18.2, SPARK=1.4.1, Docker=1.12.0 and i am using this container gelog/adam https://hub.docker.com/r/gelog/adam/

[davidonlaptop]

Hi @akhmess, adding more more memory is only going to help you if this is the bottleneck for your use case.

[akhmees]

The thing is so far samtools is outperforming ADAM on flagstat command.
Although i am running the test on ADAM file.
On Wed, Sep 28, 2016 at 11:12 PM David Lauzon [email protected]
wrote:

Hi @akhmess https://github.com/akhmess, adding more more memory is only
going to help you if this is the bottleneck for your use case.

—
You are receiving this because you were mentioned.
Reply to this email directly, view it on GitHub
#4 (comment),
or mute the thread
https://github.com/notifications/unsubscribe-auth/AOLtvz6Q-MOmwiua-HIjUcJUWUhL_mMPks5quso8gaJpZM4GRzYt
.

[aapril]

Send this comment to Frank at Berkeley

Prof April

On Sep 28, 2016, at 16:52, akhmess [email protected] wrote:

The thing is so far samtools is outperforming ADAM on flagstat command.
Although i am running the test on ADAM file.
On Wed, Sep 28, 2016 at 11:12 PM David Lauzon [email protected]
wrote:

Hi @akhmess https://github.com/akhmess, adding more more memory is only
going to help you if this is the bottleneck for your use case.

—
You are receiving this because you were mentioned.
Reply to this email directly, view it on GitHub
#4 (comment),
or mute the thread
https://github.com/notifications/unsubscribe-auth/AOLtvz6Q-MOmwiua-HIjUcJUWUhL_mMPks5quso8gaJpZM4GRzYt
.

—
You are receiving this because you are subscribed to this thread.
Reply to this email directly, view it on GitHub, or mute the thread.

[davidonlaptop]

Good point @aapril. Yes since it is a ADAM related issue, you should report it to the ADAM community.