search

tseemann / prokka

:zap: :aquarius: Rapid prokaryotic genome annotation
839 stars 226 forks source link

initialization of Vm error #557

Open nanzhen102 opened 3 years ago

nanzhen102 commented 3 years ago

Hi, I am new to prokka and am lost... when I try to test run prokka in compute canada, an error happens: ERROR: [09:32:36] Looking for 'makeblastdb' - found /home/nanzhen/prokka/bin/../binaries/linux/makeblastdb [09:32:37] Determined makeblastdb version is 002009 from 'makeblastdb: 2.9.0+' [09:32:37] Looking for 'minced' - found /home/nanzhen/prokka/bin/../binaries/linux/../common/minced [09:32:37] Coult not parse version from 'Error occurred during initialization of VM'

I used module load StdEnv/2020 bioperl/1.7.7 to run prokka.

Thank you and waiting for your help!

cmkobel commented 2 years ago

Whereas @nanzhen102's issue is that the java vm for minced can't start at all, and that the output to stdout messes with the version parsing of prokka, I'm having a similar issue. My issue isn't that I can't run minced, but that when I run minced, the java VM outputs a warning message that interferes in the same way.

[02:52:01] Looking for 'minced' - found /home/cmkobel/assemblycomparator2/conda_base/d2564f44caed87c908e089840fefdba9_/bin/minced
[02:52:02] Coult not parse version from '[0.086s][warning][os,container] Duplicate cpuset controllers detected. Picking /sys/fs/cgroup/cpuset, skipping /cgroup/cpuset.'

This issue occurs because the minced --version call in the prokka binary returns:

[0.005s][warning][os,container] Duplicate cpuset controllers detected. Picking /sys/fs/cgroup/cpuset, skipping /cgroup/cpuset.
minced 0.4.2
MinCED - Mining CRISPRs in Environmental Datasets (version 0.4.2)
Copyright 2011      Florent Angly     <florent.angly@gmail.com>
          2013-2018 Connor Skennerton <c.skennerton@gmail.com>

Minced is a modified version of CRT (www.room220.com/crt)
Charles Bland et al., CRISPR Recognition Tool (CRT): a tool for automatic
detection of clustered regularly interspaced palindromic repeats
BMC Bioinformatics 8, no. 1 (2007): 209.
Distributed under the GNU General Public License version 3

Prokka expects the version number to be on the first line, which is now the second line.

For me, this is an aperiodic error that only occurs sometimes. I'm having a hard time reproducing the error, but it seems that specific genomes are more prone to fail than others.

Hacky solutions

I found that limiting the openjdk version to max 17.0.2 solves the issue. This can be set by explicitly defining the wanted version of openjdk when installing prokka: openjdk<=17.0.2

Another solution that might work (I haven't tested it, but it seems legit) is to set some java tool options before running prokka. Thus, when minced is called (by prokka), the warning will be hidden.

JAVA_TOOL_OPTIONS="-Xlog:disable -Xlog:all=warning:stderr:uptime,level,tags"

Real solution

The only good solution to this issue is to make prokka parse the correct line of the messy output that the minced --version call makes.