Testing the Jupyter Notebooks on k8s with V4

[dwolfson]

I've created a test summary spreadsheet to track testing of V4 against the Jupyter notebooks. I'm planning on making some enhancements to indicate sequence of tests since the order seems to be quite important - not just in terms of name clashes but sometimes in terms of resource consumption. There are different tabs for testing different repository types on different days.

Note that this testing suggests some future notebook cleanup and is useful in experimenting with resource allocation,monitoring, and management with k8s environments. Testing is not showing issues in Egeria itself.

Lab Tests.numbers.zip

I'll continue testing..

[planetf1]

Thanks for the testing. I suspect these issues have been here for a long time. It's a good point that we need to ensure the lab user can explore the different labs at their leisure, so we should avoid resource clashes between them, and do enough to reset the lab at the beginning (not just the end, as they may not make it all the way through)

There are also some performance limitations - some of the behaviour is asynchronous so clicking 'too quickly' (and speed will depend on the environment's performance characteristics + choice of repo etc) could lead to failure. Here we may need retries, warnings, or a 'minimal arrival time' pauses in the script

[planetf1]

This probably belongs to the jupyer repository.... unless it's specific to the charts ... since these are functional challenges.

If we specifically hit an issue with how egeria is setup, or it's something we need to change, then let's raise an issue in the charts repo?

[planetf1]

(we can 'transfer' the issue if it makes sense to do this split)

[planetf1]

Some comments on the spread sheet with March 23 entries

• The 'working with standard models' lab on inmemory fails with null glossary usage due to there being two glossaries -- the first one is from automated curation. That does not have all the same properties, resulting in the failure. We discussed this - please open a distinct issue on the notebooks. It requires the code just before Figure.2 in that notebook to be resiliant to a missing 'usage' - or to just look for the glossary it understands.
• Automated curation with graph worked for me -- but I think there are intermittent timing bugs, and also (probably in this case) I think you're finding 'state' challenges, ie with additional repo contents breaking assumptions in the logic.
• Open lineage/graph - ok for me. Your errors may be timing
• understanding asset/graph - probably related to the first issue above, where we have a glossary without a defined usage
• xtdb - similar comment to graph on automated curation re: timing

So I agree, all look like notebook issues. Transferring.

[dwolfson]

A bit of an update. After some bug fixes and updates to the notebooks, the labs seem to be running cleanly. Longer term, there is refactoring and cleanup to do but they are currently stable.

[planetf1]

Most appeared to be timing -- but the first was a logic issue when running labs in different orders. Was this addressed? I don't see a specific issue/PR ?