Under certain circumstance, the client errors running an update module with "Could not execute update module: fork/exec /usr/share/mender/modules/v3/logger-update-module: no such file or directory".
This has been seen in CI few times, and also been able to reproduce locally by Lluis.
A log showing the issue can be found here (tip: grep for "fork/exec" and then browse the previous test case)
What happens is the following:
- The main part of test case "Continue, then fail in ArtifactInstall_Enter-case9" runs correctly.
- The client has trouble reporting status to the server, running into various "update-retry-report [none] -> update-status-report [none]" transitions
- The client gets shutdown (systemctl stop) still while retrying to send the report
- Next test case starts, "continue with higher priority-case10"
- Detects the previous error: "Mender shut down in state: update-status-report"
- Tries to run the update module:
- Calling module: /usr/share/mender/modules/v3/logger-update-module ArtifactFailure /var/lib/mender/modules/v3/payloads/0000/tree
- Could not execute update module: fork/exec /usr/share/mender/modules/v3/logger-update-module: no such file or directory
The "no such file or directory" is not really the module itself, but the chdir that the code does into the payload directory, which does not exist after Mender shutdown.
The client eventually recovers, so I am not sure if this is a real issue or just a test weakness.