CI publish_tagged_rpm stage succeeds in spite of failure to update the EOS repo

Summary

When running the update repodata script during the publish_tagged_rpms stage it might fail, but the job is marked as a success. This results hosts failing to upgrade due to missing rpms.

Example job: https://gitlab.cern.ch/cta/CTA/-/jobs/25712951 Towards the end it shows the following message:

Cannot access/write repodata files: [Errno 2] No such file or directory

Steps to reproduce

Undefined, the access/write error may be due to a transient network/EOS problem.

What is the current bug behaviour?

The repomd.xml file is not updated with the information for the new rpms.

[rbachman@aiadm81 ~]$ ls -l /eos/user/c/ctareg/www/cta-repo/repodata/
total 4324
-rw-r--r--. 1 ctareg def-cg  262780 Oct 26 16:25 162901623773ba642b347545d78fd22bdb6f809cf5abecbb7ebb99eb4f218abb-primary.xml.gz
-rw-r--r--. 1 ctareg def-cg 1155126 Oct 26 16:25 487d4e85515e23cc9e00bfd790d3e67591f857f7d8b7acc0253b1569791ad689-primary.sqlite.bz2
-rw-r--r--. 1 ctareg def-cg  394590 Oct 26 16:25 7440312bb0120ca9351bed902268e3133d7533f4f08c3dd1d22ffc2c4b6f7191-other.sqlite.bz2
-rw-r--r--. 1 ctareg def-cg  944539 Oct 26 16:25 80cee4f50e0677bb2f64d3c123184cd791269cf5efd4597033907bddbfdf6405-filelists.sqlite.bz2
-rw-r--r--. 1 ctareg def-cg  301479 Oct 26 16:25 cb8ce8298100b66337a16a5b0b0b298371325b1b8739920c8ceedf69e581a0f9-other.xml.gz
-rw-r--r--. 1 ctareg def-cg 1364362 Oct 26 16:25 d4b921b5286f2710283ade13995e2089c2c5f9daaf8e809258772e39e09daf63-filelists.xml.gz
-rw-r--r--. 1 ctareg def-cg    3254 Oct 26 16:25 repomd.xml

What is the expected correct behaviour?

The repomd.xml file is updated with the current rpm data.

Relevant logs and/or screenshots

Possible causes

Transient EOS problems, network issues, ...