QA (public) » openQA Project (public)

Observation¶

openQA test in scenario sle-15-SP5-Online-s390x-xfstests_xfs-generic@s390x-kvm-sle15 fails in
generate_report
All xfstests runs in sle-15-SP5 s390x fails on that issue.

In this specific case the connection attempt with failed curl was from (reading out from vars.json)
"SUT_IP" : "s390kvm082.suse.de",
"VIRSH_GUEST" : "10.161.145.82",
"VIRSH_HOSTNAME" : "s390zp18.suse.de",

At first, I thought this is the same issue under debugging in #120261, but after that solution(https://github.com/os-autoinst/openQA/pull/4935/files) merged our fails in s390x still. By looking into the details I don't know why these tests still use worker2.oqa.suse.de as the download IP. Previous last good used IP address not use FQDN. May need some help by the tools team.

okurz ran time curl -O http://worker2.oqa.suse.de:20343/rfhqRYw7W_g045X2/files/status.log which reproduces the problem quite explicitly:

# time curl -O http://worker2.oqa.suse.de:20343/rfhqRYw7W_g045X2/files/status.log
  % Total    % Received % Xferd  Average Speed   Time    Time     Time  Current
                                 Dload  Upload   Total   Spent    Left  Speed
  0     0    0     0    0     0      0      0 --:--:--  0:02:10 --:--:--     0curl: (7) Failed to connect to worker2.oqa.suse.de port 20343: Connection timed out
real    2m11.316s

so very likely the firewall for the .oqa.suse.de zone just drops packets from 10.161.0.0

Reproducible¶

Fails since (at least) Build 40.1

Expected result¶

Last good: build38.1 http://openqa.suse.de/tests/9886322#step/generate_report/2

Suggestions¶

1. Ask SUSE-IT network admins to REJECT packets instead of DROP so that we get more clear results #122653
2. Ask SUSE-IT network admins to not block this traffic which we need for tests #122656
3. As it looks like default connect timeout for curl resolves to 2m10s (see above) so that is above our default timeouts for script_run, etc., so find a combination where curl has a chance to provide a proper error earlier. Consider using upload_logs in this specific example but this does not completely help. upload_logs uses a default timeout of 90s which is higher than the default for script_run of 30s which is still below the default for curl accounting to 2m10s. Maybe we add the parameter --connect-timeout 20 to curl or bump the timeout for upload_logs #122659
4. Ensure the original problem is fixed #122539

Further details¶

Link to latest