R: install a working R 4.4.2 using JDK 21

[balston]

R 4.4.2 with Java JDK 21 instead of JDK 8. This should be more compatible with running the Nextflow pipeline and R together eg see closed ticket 1393718

Needed first on Myriad

[balston]

A build script for R 4.4.2 and recommended packages has been done which uses the java/21.0.4 module:

https://github.com/UCL-RITS/rcps-buildscripts/blob/f5ffc0b053aaa744110947da3c9bc8facdedfe7a/R-4.4.2_install

and run on Myriad.

[balston]

The additional packages build script R-4.4.2_packages_install is prepared but I've been going round in circles trying to get a working set of the external geographic tools to work together PROJ.4, GDAL, GMT and GEOS. I think I may finally have got them working with the following:

• PROJ.4 9.2.0
  build script PROJ.4-9.2.0_install
  https://proj.org/en/stable/
• GDAL 3.3.3
  build script gdal-3.3.3_install
  https://gdal.org/en/stable/
• GMT 6.5.0
  build script gmt-6.5.0_install
  https://www.generic-mapping-tools.org/
• GEOS 3.9.1
  build script geos-3.9.1_install
  https://libgeos.org/

[balston]

Now trying to run the R-4.4.2_packages_install from a job as it takes some time to complete. Job script is in:

~ccspapp/Software/R/build-R-4.4.2-packages.sh

[balston]

Job is now running:

job-ID  prior   name       user         state submit/start at     queue                          slots ja-task-ID
-----------------------------------------------------------------------------------------------------------------
 460617 3.50000 R-4.4.2-pa ccspapp      r     11/27/2024 18:03:08 bran@node-d00a-026                 2

[balston]

Job finished overnight:

jobhist
        FSTIME        |       FETIME        |   HOSTNAME    |  OWNER  | JOB NUMBER | TASK NUMBER | EXIT STATUS |        JOB NAME
----------------------+---------------------+---------------+---------+------------+-------------+-------------+-------------------------
  2024-11-27 18:05:11 | 2024-11-27 22:34:33 | node-d00a-026 | ccspapp |     460617 |           0 |           0 | R-4.4.2-packages-build

Now checking how much worked.

[balston]

building package V8:

https://cran.r-project.org/package=V8

fails.

[balston]

I have a comment:

# Can no longer use external V8 - must set DOWNLOAD_STATIC_LIBV8=1 before 
# installling the V8 package

maybe this is no longer available/working for RedHat 7. Checking

[balston]

It looks like an older version of the V8 package may work. Trying:

packageurl <- "https://cran.r-project.org/src/contrib/Archive/V8/V8_5.0.1.tar.gz"
install.packages(packageurl, repos=NULL, type="source")

[balston]

This does seem to work.

[balston]

ok now I'm going to start again after moving /shared/ucl/apps/R/R-4.4.2-OpenBLAS out of the way so starting a fresh build. First:

./R-4.4.2_install 2>&1 | tee ~/Software/R/R-4.4.2_install .log-2

[balston]

base R and recommended packages re-built. Just running some of the demos to test.

[balston]

submitted additional packages build job from ~ccspapp/Software/R:

qsub build-R-4.4.2-packages.sh
job-ID  prior   name       user         state submit/start at     queue                          slots ja-task-ID
-----------------------------------------------------------------------------------------------------------------
 468436 0.00000 R-4.4.2-pa ccspapp      qw    11/29/2024 11:48:02                                    2

[balston]

Build job has finished with no obvious errors. Log is in /home/ccspapp/Scratch/R/R-4.4.2-packages-build.o468436

[balston]

Installing R MPI support packages like doMPI:

./R-4.4.2_MPI_install 2>&1 | tee ~/Software/R/R-4.4.2_MPI_install.log

[balston]

looks ok. Next step is to install the R 4.4.2 module file and module bundle.

[balston]

I've run some tests on the Myriad build:

• a serial benchmakeme run https://cran.r-project.org/package=benchmarkme
• an MPI test using the doMPI package https://cran.r-project.org/package=doMPI

the tests have worked ok so R 4.4.2 is now ready to install on all clusters. Will do Kathleen next.

[balston]

On Kathleen I'm running the build scripts in this order:

module unload gcc-libs
./R-4.4.2_install 2>&1 | tee ~/Software/R/R-4.4.2_install.log
./PROJ.4-9.2.0_install 2>&1 | tee ~/Software/PROJ.4/PROJ.4-9.2.0_install.log
./gdal-3.3.3_install 2>&1 | tee ~/Software/GDAL/gdal-3.3.3_install.log
./gmt-6.5.0_install 2>&1 | tee ~/Software/GMT/gmt-6.5.0_install.log

GEOS 3.9.1 is already installed on Kathleen.

[balston]

R 4.4.2 additional packages build job located here on Kathleen:

~ccspapp/Software/r/build-R-4.4.2-packages.sh

submitted earlier.

qstat
job-ID  prior   name       user         state submit/start at     queue                          slots ja-task-ID
-----------------------------------------------------------------------------------------------------------------
 522923 3.50000 R-4.4.2-pa ccspapp      r     12/05/2024 14:31:11 [email protected]    80

[balston]

I've checked the log for the additional packages build job and it looks as everything. installed ok.

[balston]

now running the MPI support packages installer script:

./R-4.4.2_MPI_install 2>&1 | tee ~/Software/R/R-4.4.2_MPI_install.log

[balston]

I've run a test job on Kathleen using the doMPI package and this has worked correctly. Will now install on all clusters:

• kathleen;
• michael;
• myriad;
• young