Open wodanaz opened 3 years ago
The following error is because someone has exhausted disk space on the node:
Caused by: java.io.IOException: No space left on device
In this case I think it's the /tmp
directory has filled up on x2-05-3.genome.duke.edu and your program cannot continue.
Send an email to gcb-help@duke.edu so the sysadmins can take a look at this node for what used up all the space.
that makes sense.
Thanks, I just put a ticket
Thank you!
If you need to get going immediately you could try avoiding this messed up node.
There is a --exclude=<node name list>
flag for sbatch, but I haven't used it.
You could try adding --exclude=x2-05-3
to the sbatch invocations.
For example to add the flag to all the array jobs we run you could add it to the following line:
Something like:
sbatch --exclude=x2-05-3 --wait --array=1...
I am trying this now. I will let you know. I have not use it before either
I think I solved for picard and for gatk. There was an issue with gatk as well and solved by adding "-Djava.io.tmpdir=/data/covid19lab/tmp" to the lines with gatk
Hard coding the tmp directory like this will hurt portability of the pipeline.
We already have a temporary directory that is cleaned up by the pipeline: $EVDIR
.
@wodanaz Are you good with me changing the scripts to store the tmp files within $EVDIR
?
That is true, Absolutely. I am okay with that. $EVDIR is actually simpler.
Also, where would be the best place to put the code for the account '-A covid19lab' ?
There is an environment variable SBATCH_ACCOUNT that can be exported to apply this setting to all sbatch calls. Right now you could turn this on by running the following in your HARDAC terminal session:
export SBATCH_ACCOUNT=covid19lab
Then within that terminal session when you run-escape-variants.sh or run-dds-escape-variants.sh the account will be used.
Since this export step is really easy to forget I would like to make a change to put the account name in a config file. That way you could set it up once. How does that sound @wodanaz?
It sounds like a great plan. I am afraid I will start forgetting to switch if I need to use wraycompute or mcclaylab accounts
Hi John, hope you are well.
I am not sure why markduplicates is not working well anymore. I thought it was a memory allocation issue but not. I tried allocating more memory and still have the same issue. I also tried to update
Here is the output of one of the errors.