[Halld-offline] Alma9 Updates: version_5.14.2.xml

[Alexander Austregesilo]

Dear Colleagues,

There were multiple reports that mcsmear and hdgeant4 were running 
extremely slow or even hanging with the latest software release 5.14.1. 
This was traced back to the CCDB version 1.07.00.

We reverted to the previous CCDB release and implemented only minor 
changes for compatibility with python3 in CCDB version 1.06.10. As a 
result, the new software stack version_5.14.2.xml was built on 
CentOS7/RHEL7 and AlmaLinux9, as well as their container counterparts. I 
switched the default version at JLab now.

Best regards,

Alex



On 1/2/24 16:25, Alexander Austregesilo via Halld-offline wrote:
> Dear Colleagues,
>
> I noticed over the holidays that hdgeant4 was not correctly building 
> with the version set 5.14.0 on RHEL7/CentOS7. For that reason, the 
> default version set at JLab which is used for nightly builds and pull 
> request tests was reverted to 5.13.0.
>
> With few modifications to the diracxx library and the build_scripts 
> system, we now have a functioning build of the full GlueX software 
> stack on CentOS7/RHEL7 and AlmaLinux9. The default version will switch 
> over to version_5.14.1.xml before midnight tonight.
>
> Wishing a happy new year to you all,
>
> Alex
>
>
> On 12/21/2023 2:57 PM, Alexander Austregesilo via Halld-offline wrote:
>> Dear Colleagues,
>>
>> As many of you are already aware, the JLab farm has started the 
>> process of changing its operating system. The target is Alma Linux 9, 
>> an open-source distribution that is binary compatible to Red Hat 
>> (RHEL). An interactive node, "ifarm9", is already available for tests 
>> and slurm will be upgraded in January in order to be able to select 
>> Alma9 farm nodes for batch jobs.
>>
>> On our side, an upgrade of the operating system comes with several 
>> advantages. With a more recent compiler version (gcc11), we will be 
>> able to upgrade root and geant4 among other packages which in turns 
>> allows us to use modern features and keep up with developments. On 
>> the other hand, a few older dependencies will be deprecated, most 
>> notably python2.
>>
>> In the recent weeks, numerous issues with our current software stack 
>> and build system were identified and fixed. Big thanks to Sean and 
>> Richard for paving the way, and a few others who helped testing.
>>
>> We released a new version set version_5.14.0.xml which has minimal 
>> changes to its predecessor, but was built successfully on AlmaLinux9. 
>> From now on, you can set up your environment on ifarm9 with the usual 
>> commands:
>>
>>> source /group/halld/Software/build_scripts/gluex_env_boot_jlab.sh
>>>
>>> gxenv /group/halld/www/halldweb/html/halld_versions/version_5.14.0.xml
>>>
>> I have successfully processed one evio file. The produced REST file 
>> is compatible with the results on CentOS7. The same is true for 
>> analysis (PART) trees, even though the ReactionFilter plugin may 
>> crash at the very end of the input file. This problem was documented 
>> in this issue on github:
>>
>> https://github.com/JeffersonLab/halld_recon/issues/613
>>
>> MCWrapper and its dependency on python2-based RCDB prevented the 
>> testing of the full Monte Carlo Simulation chain so far. The new 
>> version set should be helpful to debug the workflow.
>>
>> Feel free to try it out. We would appreciate your feedback, 
>> especially if you are the maintainer of programs, plugins or scripts. 
>> We also plan to release a singularity/apptainer container that is 
>> compatible with this build.
>>
>> Best regards and happy holidays,
>>
>> Alex
>>
-- 
Alexander Austregesilo

Staff Scientist - Experimental Nuclear Physics
Thomas Jefferson National Accelerator Facility
Newport News, VA
aaustreg at jlab.org
(757) 269-6982