These are chat archives for nextflow-io/nextflow

21st
Dec 2016
Félix C. Morency
@fmorency
Dec 21 2016 14:47
very nice
Félix C. Morency
@fmorency
Dec 21 2016 19:43
Caused by:
  com.esotericsoftware.kryo.KryoException: java.lang.ArrayIndexOutOfBoundsException: -2
Serialization trace:
containerMemory (nextflow.processor.TaskBean)
any idea what might be causing this?
Paolo Di Tommaso
@pditommaso
Dec 21 2016 19:45
weird
please include the full stack trace (via pastebin.com or similar)
Félix C. Morency
@fmorency
Dec 21 2016 19:45
ok it works after I updated the other nextflow to 0.23
i was running 0.22.5 on computer1 and 0.23 on computer2
it works with both computers on 0.23
Paolo Di Tommaso
@pditommaso
Dec 21 2016 19:51
when happen the error ?
using -resume ?
Félix C. Morency
@fmorency
Dec 21 2016 19:51
using -resume on a fresh run
Paolo Di Tommaso
@pditommaso
Dec 21 2016 19:52
need to see the stack trace
Félix C. Morency
@fmorency
Dec 21 2016 19:54
damn I deleted the .nextflow.log file
Paolo Di Tommaso
@pditommaso
Dec 21 2016 19:54
ok, no issue any more :)
Mike Smoot
@mes5k
Dec 21 2016 20:40
Hi Paolo, I'm experimenting with cfncluster and nextflow. I think I've got cfncluster configured correctly (nextflow, docker, etc. installed). I'm now trying to run a pipeline. Nextflow correctly submits job, and I see them in SGE's queue, but I'm not seeing any evidence that any of the worker nodes are actually doing any work. Do you have any tips for debugging an SGE cluster?
indeed qstat -s r shows nothing running.
Paolo Di Tommaso
@pditommaso
Dec 21 2016 20:59
umm, no
what's the benefit of using cfncluster instead of a NF cluster ?
Mike Smoot
@mes5k
Dec 21 2016 21:01
To run a proper job schedule that can handle multiple invocations of nextflow at once. Something you recommended.
Paolo Di Tommaso
@pditommaso
Dec 21 2016 21:02
I see, I was not remembering that
Mike Smoot
@mes5k
Dec 21 2016 21:03
I'd also eventually like the ability to specify queues for specific types of jobs, specifically to run blast on machines with a enough ram and with the database cached.
Which is not to say that I'll actually use cfncluster - I'm still just experimenting!
Paolo Di Tommaso
@pditommaso
Dec 21 2016 21:04
:+1:
Mike Smoot
@mes5k
Dec 21 2016 21:05
It looks like my worker nodes are stuck in state 'au', which I gather is not good. Will begin googling... :)
Paolo Di Tommaso
@pditommaso
Dec 21 2016 21:06
no clue what au stand for
Mike Smoot
@mes5k
Dec 21 2016 21:08
a is alarm and u is unreachable, so clearly my cluster is not correctly configured. Although the auto-scaling does work as it keeps spawning new worker nodes... sigh
Paolo Di Tommaso
@pditommaso
Dec 21 2016 21:09
:)
good business model for AWS
Mike Smoot
@mes5k
Dec 21 2016 21:10
thankfully I'm still experimenting with t2.micros, so they're not getting that much