Changes for el9-based cmssw versions

[anpicci]

I have slightly modified the scripts in order to properly handle the CMSSW sandboxes that require el9 architecture

[btovar]

@anpicci Note that with this branch we are trying to move away from parrot. The images from runos should have given you cvmfs, isn't that the case?

[anpicci]

@btovar yes, it still relies on parrot

[bryates]

@anpicci @btovar what's the status of this branch? I'm looking to run some Run 3 stuff now.

[anpicci]

@bryates , this branch is what I use for running the skimmed samples for run3

[Andrew42]

@anpicci running with your changes I came across a crash caused by a byte-string getting introduced to data which then gets serialized into a json output file. Specifically, I had to change this line to

checksum = stdout.split()[-2].decode('utf-8')

[Andrew42]

In addition to the byte-string bug, I also had to modify task.py LL71-73 to exclude this check for CMSSW_12_4_X. The XrdAdaptor::XrdStatisticsService module appears to not exist in this CMSSW release. I simply removed/commented the process.add_(xrdstats) to keep it from being added to the CMS executable.

Without this change the lobster task will make it through set up, but crash once it attempts to do the cmsRun executable of the form:

>>> executing 'cmsRun -j report.xml test_LHE_cfg_mod.py' @ Fri Oct  3 17:04:24 2025 
>> using /tmp/tmpmgxbm8kh to store command output
>> cmd: ----- Begin Fatal Exception 03-Oct-2025 17:04:28 EDT-----------------------
>> cmd: An exception of category 'PluginNotFound' occurred while
>> cmd: [0] Constructing the EventProcessor
>> cmd: Exception Message:
>> cmd: Unable to find plugin 'XrdAdaptor::XrdStatisticsService' in category 'CMS EDM Framework Service'. Please check spelling of name.
>> cmd: ----- End Fatal Exception -------------------------------------------------

Removing the module did not seem to cause any other issues for the task, though likely results in some missing monitoring metrics.

[Andrew42]

@anpicci is this change missing an 'and'?

[Andrew42]

@anpicci another thing that I needed to change was the ndcctools version that gets installed in the conda environment.

Right now the yaml file has it pinned to 7.11.1, however, in order to use the --runos al8-wq-7.15.9 option for the WQ factory, @btovar said that I should use 7.15.9 or newer. The version of ndcctools I updated to in my working environment was 7.15.11, so we can either use that or pin it to 7.15.9.

[bryates]

@ywan2 this was helpful for debugging, but I don't know if we need it in the official release. I guess it doesn't hurt. What do you think @Andrew42 @anpicci?

[Andrew42]

I agree that we should keep any "mandatory" print statements to a minimum. I believe this would only ever get printed once when the user does a lobster process path/to/lobster_config.py, so probably not very disruptive, but maybe there should be a better way to log/convey this information to the user.

[bryates]

What is biglib used for? We did our full Run 3 production without it.