When running 8.2 beta precompiled: new message (not in 8.1)
cat runPBcR.sge.out.02
-bash: line 10: /var/lib/gridengine/ebb-default/common/settings.sh: No such file or directory
Error: spec file /global/group/home/lmall_01/projects/Bt/assembly/247/v_9//tempBt247_Subreads/Bt247_Subreads.spec does not exist. Double-check your paths and try again.
due to this script: runPBcR.sge.out.02.sh
#!/bin/bash
#
if [ "x$SGE_ROOT" != "x" ]; then
# Attempt to (re)configure SGE. For reasons Bri doesn't know,
# jobs submitted to SGE, and running under SGE, fail to read his
# .tcshrc (or .bashrc, limited testing), and so they don't setup
# SGE (or ANY other paths, etc) properly. For the record,
# interactive SGE logins (qlogin, etc) DO set the environment.
. $SGE_ROOT/$SGE_CELL/common/settings.sh
fi
Comments:
$SGE_ROOT/$SGE_CELL/common/settings.sh
is not present in all GE versions, making the jobs crash if missing. Mine, GridEngine 6.2u5-7.1 is missing it for example.
sourcing .tcshrc or .bashrc can be a multifactorial problem, commons reasons are invalid or incoherent uid across the system, permissions, /etc/profile or /etc/bash.bashrc empty or missing the sourcing or referencing each other, PATH declaration or SGE variables initialisation.
The problem can also come form the queue "shell start mode" behaviour. "posix_compliant" and "unix_behaviour" have given me headaches, "script_from_stdin" simulates something like the interactive mode and the env in sourced.
qconf -mconf
shell_start_mode script_from_stdin
qconf -mq all.q
shell_start_mode script_from_stdin
I add -V (export whole env) in the SGE stanza of the spec file and force a reload of bash (which then source the proper files as it thinks it is opened in interactive mode.)
sge = -V -S /bin/bash -l mem_free=4G -r y -wd $assembly_path
sgeScript = -V -pe smp 1 -l mem_free=4G -S /bin/bash -r y -wd $assembly_path
sgeOverlap = -V -pe smp 4 -l mem_free=2G -S /bin/bash -r y -wd $assembly_path
sgeFragmentCorrection = -V -pe smp 4 -l mem_free=5G -S /bin/bash -r y -wd $assembly_path
sgeOverlapCorrection = -V -pe smp 4 -l mem_free=3G -S /bin/bash -r y -wd $assembly_path
sgeConsensus = -V -pe smp 1 -S /bin/bash -r y -wd $assembly_path
Second error:
I don't know whether is comes from the missing config.sh, but the spec file is not copied in the temp subdirectory.
Hi,
I am not sure why you are only seeing the settings.sh error from CA 8.2 as CA 8.1 has been sourcing the $SGE_ROOT/$SGE_CELL/common/settings.sh script as well. As long as the login nodes will be able to run the SGE commands (qsub/qstat/etc), your run should work even without this script.
As far as your second error, can your provide the full output of the pipeline (from the start before it submits to grid and all grid output) as well as your command line run? The spec file gets copied to the directory before grid submission.
As
Hi,
Thanks for your answer!
unfortunately, I erased the output files, and only archived the temp
directory... which doesn't contain more info. I guess the config.sh
problem is minor then, I used the preassembled reads from the smrt
pipeline (HGAP quality mode) instead of directly calling pbcr. I would
still be interested by comparing the results to a manual pbcr run.
I give it another shot.
On 02/12/14 21:27, Sergey Koren wrote:
Related
Bugs: #287
Hi,
If I understand you correctly, you were providing the pre-assembled HGAP reads to PBcR? It should not cause the spec file error but it will not work properly as PBcR is designed to take the raw uncorrected filtered sequences. You can assemble the pre-assembled reads with CA but this is what HGAP does already.