QA (public) » openQA Project (public) » openQA Infrastructure (public)

Observation¶

HanaSR test in Multi-Machine configuration started to fail recently while trying to reach gitlab.suse.de:
wmp_check_process

Tools Team was contacted over slack and a joint debug session was done. Some details:

• SUTs (qemu_ppc64le VMs in diesel and petrol) running in Multi-Machine setup, fail to download assets from gitlab.suse.de.
• Command is: curl -s -k https://gitlab.suse.de/qa-css/wmp_basic_tests/-/archive/master/wmp_basic_tests-master.tgz -o-
• Same command from the worker (diesel or petrol) works
• Taking control of the jobs it was possible to see that SUTs are able to resolve gitlab.suse.de to both its IPv4 and IPv6 addresses
• ICMP from the SUTs to gitlab.suse.de over IPv4 fail with a timeout.
• ICMP from the SUTs to gitlab.suse.de over IPv6 fail with Network is Unreachable. (IPv6 is not configured in the SUTs)
• IP addresses of the SUTs are 10.0.2.15 and 10.0.2.16
• Name server in both nodes is 10.0.2.1 (IP address of the support server running in parallel to the jobs)
• After checking with tcpdump in the worker, 2 extra addresses were identified: 10.1.13.76 (related to the node with IP address 10.0.2.15) and 10.1.13.78 (related to the node with IP address 10.0.2.16)
• Same scenario is working in qemu_x86_64.
• After modifying the firewall rules in the worker with ip rule add from 10.1.13.78 table nowg prio 1 and ip rule add from 10.1.13.76 table nowg prio 1, SUTs were able to ping gitlab.suse.de and download assets from it. Tests passed: https://openqa.suse.de/tests/16816617 & https://openqa.suse.de/tests/16816618

Reproducible¶

Fails since (at least) Build 119.6

Expected result¶

Last good: 117.1 (or more recent)

Further details¶

Always latest result in this scenario: latest