Ensuring asset files integrity (was: "An error occurred during the installation" on images)
|Target version:||QA - future|
As a test reviewer I want to have checksums of asset files so that I can triage infrastructure/product/test issues in case the media seem to be corrupt
- AC1: The checksum of asset files is reported when configured
- AC2: No checksum checks of ISO files in general (too expensive)
- more checksum checks and such, maybe on the level of tests, not os-autoinst or openQA
- don't sync images to o.s.d when rsync.pl can already know it's broken (or incomplete)
There is a bug open in bugzilla bsc#994617 for the same symptoms and the bug reviewers judge it's an openQA issue. The issue is happening from time to time:
#4 Updated by coolo over 3 years ago
- Status changed from New to Feedback
The same ISO worked fine here: https://openqa.suse.de/tests/569770 - so if the ARM worker has a problem with its NFS implementation, you better report that in bugzilla related to the host system. Does this happen on other workers as well?
#5 Updated by okurz over 3 years ago
- Description updated (diff)
It does not look like we have seen this error on other workers so far. As can be seen in the description this issue was originally reported in bugzilla and we are playing ping-pong here. I should have kept the bugzilla bug open and just change the product and component to refer to test infrastructure I guess. I reopened bsc#994617 with the new information. Still, should we add more checksums or maybe we already have more checksums now or not on the used ISOs?
#6 Updated by coolo about 3 years ago
- Status changed from Feedback to New
- Priority changed from High to Low
I don't want checksums of ISOs in general - it's pretty expensive and it would only be a measure to debug NFS implementations. We can potentionally add a workers.ini setting, but not on high prio
#14 Updated by rpalethorpe about 1 month ago
This would be very helpful in determining where problems happen:
Basically the following could be done:
1) Take sha256 hash of assets on the worker (maybe just for problem assets), create a sum file
2) Check the hash after uploading and every other operation where an asset is moved or copied.
From what I can see there have been many bugs with the asset handling.