Hi, I am getting an error while running my pipelin...
# random
s
Hi, I am getting an error while running my pipeline on the cluster. I tried to search about this issue, but there isn't much info. about it
Unexpected error while dumping DAG
🤔
p
Hi! Any chance you could provide a bit moreinfo please?
s
To help we'd love a few more details please: 1. The Nextflow command you used to launch the pipeline 2. Any custom Nextflow config files 3. The full error message 4. The corresponding
.nextflow.log
file, where you are running the analysis. Remember to use triple backslashes on their own lines for code blocks. Also take a look at https://nf-co.re/usage/troubleshooting for more information on debugging failed pipelines. Thanks!
t
The error is
Session aborted -- Cause: SIGINT
. SIGINT is the user-generated signal normally created by pressing
Ctrl-C
. Did you (or your cat cattyping) do that? If not, is this reproducible?
➕ 2
a
Is that the whole error? Sometimes the actual issue is above what you've posted. SIGINT can be the result of user (like Thomas said) or a program erroring.
party this 1
s
Thanks for all the inputs🙌🏻, turns out nextflow generates the
SIGINT
error when the disk space is full. After looking deeper into the issue, and running with multiple datasets, I observed that the pipeline failed only when the raw data was huge in size and used up all of the disk space. We have a disk space limit for each user for the cluster usage at our institute🫠
👍 2
I am currently looking at the
publishDir
mode set to copy but will try this with symlink as xeniumranger makes changes to the input files inplace
t
Interesting, good to know! I would not change the
publishDir
mode, because that makes links in your output to the
work
directory (and if you wipe that, your output is gone as well). I think it is preferable to clean up
work
(with
nextflow clean
or
rm -r
) after each run. If you can run the samples separately and clean up after each run, that may help. Alternatively, there is the nf-core/boost plugin, which can clean up temporary file during the run, but then you can't use
-resume
anymore. If you have a large scratch space, you can also put the work directory there.
👀 1
party this 1
If
work
and the output directory are on the same disk, you could set the
publishDir
mode to
link
, so it will create hard links. This is much safer, as the file continues to exist as long as there is at least on link to it, and even with multiple link there is only one copy of the file taking up space. (This could be a viable option is the files that are actually being published are very large.)
party this 2
s
Thanks @Thomas Danhorn
👍 1