Debian: coredump files not generated in container regression tc test_coredump_file_creation

[Jiping Ma]

Debian: coredumps unexpectedly not getting generated by tc
functional/z_containers/test_core_dump_creation.py::test_coredump_file_creation

Severity

standard

Steps to Reproduce

Run jenkins tc that kills process and looks for coredump files to be generated
/z_containers/test_core_dump_creation.py::test_coredump_file_creation

Repeats Kill the process operations
kubectl exec ng -n default – /bin/sh -c "kill -6 <processid>"
Then tries to find the coredump
'ls -l /var/lib/systemd/coredump | grep core.nginx.*.xz'

Expected Behavior

Expect that coredump file(s) will be generated
Actual Behavior

No coredump files were actually generated
E utils.exceptions.SSHExecCommandFailed: Failed to execute command via SSH.
E Details: Non-zero return code for cmd: ls -l /var/lib/systemd/coredump | grep core.nginx.*.xz. Output:

utils/clients/ssh.py:574: SSHExecCommandFailed

Reproducibility

yes

System Configuration+

simplex Debian
jn
Last Pass
The tc get further on centos (ie. actually creates the coredump files) but fails on the teardown

Timestamp/Logs
Lab: r730_1

System Type: AIO-SX
OpenStack Deployed: False
Ends at: 20220427 00:06:00
Session Tag: None
Session ID: 4789016f-d970-428f-96b9-3346273f2333
Software Version: 22.02

Alarms

Please indicate if there are any alarms observed.

If there are any alarms please list them here

Test Activity

Regression Testing (Debian)

Workaround

Describe workaround if available

Attachments

[OpenStack Infra (hudson-openstack)]

Fix proposed to branch: master
Review: https://review.opendev.org/c/starlingx/utilities/+/841090

[OpenStack Infra (hudson-openstack)]

Reviewed: https://review.opendev.org/c/starlingx/utilities/+/841090
Committed: https://opendev.org/starlingx/utilities/commit/7d71d83678c58dcff63e6f02ce67647b67cbd3ab
Submitter: "Zuul (22348)"
Branch: master

commit 7d71d83678c58dcff63e6f02ce67647b67cbd3ab
Author: Jiping Ma <email address hidden>
Date: Sun May 8 18:04:12 2022 -0700

    Debian: Fix the issue coredump files can not be generated.

    Compares with v219 systemd(CentOS use this one), v247 systemd
    adds the parameter META_ARGV_HOSTNAME as the upstream commit:
    https://github.com/systemd/systemd/commit/f45b8015513d38ee5f7cc361db9c5b88c9aae704
    which was integrated in the v238 systemd that need be 7 arguments.
    But they are 6 arguments in our old config file, so there prints
    out the error information in kernel dmesg that caused coredump
    file can not be generated.

    systemd-coredump[1496]: Not enough arguments passed by the kernel
    (6, expected 7).

    This commit adds coredump-sysctl-debian.conf for debian that adds
    parameters as the code in src/coredump/coredump.c.
    https://github.com/systemd/systemd/blob/main/src/coredump/coredump.c.

    META_ARGV_PID, /* %P: as seen in the initial pid namespace */
    META_ARGV_UID, /* %u: as seen in the initial user namespace */
    META_ARGV_GID, /* %g: as seen in the initial user namespace */
    META_ARGV_SIGNAL, /* %s: number of signal causing dump */
    META_ARGV_TIMESTAMP, /* %t: time of dump, expressed as
    seconds since the Epoch (we expand this to µs granularity) */
    META_ARGV_RLIMIT, /* %c: core file size soft resource limit */
    META_ARGV_HOSTNAME,/* %h: hostname */
    _META_ARGV_MAX,

    The coredump-sysctl-debian.conf also includes default settings for
    enabling coredumps by default, as expected by StarlingX.

    Test Plan:
    PASS: Build debian ISO and boot up with qemu.
    PASS: Generate the coredump file after the fix.
       1. sleep 10 &
       2. killall -SIGSEGV sleep
       3. ls /var/lib/systemd/coredump/
        core.sleep.1878.df5133eec12a4ee2a158237c7b5b41a8.1746.1653040470000000.zst

    PASS: Build CentOS ISO and boot up in the lab.
    PASS: Generate the coredump file.

    Closes-bug: 1972135

    Signed-off-by: Jiping Ma <email address hidden>
    Change-Id: I5c9a4d820430d6cad51636cd94b3363fadef4748