QA » openQA Project » openQA Infrastructure

Just for the record, it seems that the create_hdd_textmode job really created and uploaded the image (opensuse-Tumbleweed-aarch64-20201028-textmode@aarch64.qcow2: Processing chunk 900/900, avg. speed ~406.312 KiB/s). So an issue here can be outruled.

I doubt that this is an aarch64 specific issue. Our cleanup code considers the architectures equally. Of course the size limit of the "openSUSE Tumbleweed AArch64" job group might not be big enough. However, the cleanup algorithm actually shouldn't consider assets of pending jobs (state is not "done" or "cancelled") at all anyways. I've also double checked some time ago whether that's actually the case when a similar issue came up. Nevertheless the asset has been removed by the cleanup algorithm:

martchus@ariel:~> xzgrep opensuse-Tumbleweed-aarch64-20201028-textmode@aarch64.qcow2 /var/log/openqa_gru.2.xz
[2020-10-29T03:00:12.0988 UTC] [debug] [pid:28083] Checking whether asset hdd/opensuse-Tumbleweed-aarch64-20201028-textmode@aarch64.qcow2 (899416064) fits into group 3 (418768782)
[2020-10-29T03:00:16.0400 UTC] [info] [pid:28083] Removing asset hdd/opensuse-Tumbleweed-aarch64-20201028-textmode@aarch64.qcow2 (belonging to job groups: 3)
[2020-10-29T03:00:16.0421 UTC] [info] [pid:28083] GRU: removed /var/lib/openqa/share/factory/hdd/opensuse-Tumbleweed-aarch64-20201028-textmode@aarch64.qcow2

(Job group 3 is "openSUSE Tumbleweed AArch64".)

So there's either a bug in the cleanup algorithm after all or the asset hasn't been correctly associated with the pending jobs.