Major There are a handle of "fragile" tests that are failing and impacting the CI work that hari kodungallur is working on. This ticket is to address these issues.
| For Gerrit Dashboard: NCBC-1218 | ||||||
|---|---|---|---|---|---|---|
| # | Subject | Branch | Project | Status | CR | V |
| 68587,3 | NCBC-1218: Fix broken tests impacting CI | master | couchbase-net-client | Status: MERGED | +2 | +1 |
| 68613,2 | create authenticated bucket and also create index on default | master | build | Status: MERGED | +2 | +1 |
Yeah, FTS is still dev preview, but the test should work if the indexes are set up correctly. Remember, this is a FTS index, not a GSI/N1QL index!
Its a type mapping for "landmark" documents and indexes the "content" field. Watch this video: https://www.youtube.com/watch?v=MT5YGnb80QQ
Its identical except for the bucket and type mappings.
Cool.. I created FTS index and that test now works. It looks like now we just have the bucket flush test failure, which for the last 3 builds – fail, pass, fail.
Edit: I was just trying to read/understand the code. It looks like the tests create a bucket for the bucket flush test. So likely nothing to do with the cluster setup itself. I go to the UI and look at the Logs tab and I see this:
Flushing "BucketManagerTests" failed with error:
|
{error,{prepare_flush_failed,[],[{'ns_1@172.31.5.143',bad_status}]}} (repeated 3 times)
|
I have no idea what that means.
Thanks Hari. The bucket flush test can probably ignored - its part of the mgmt API, which seldom changes if ever.
The jenkins job can't skip / ignore the tests. For now, I set the result such that if the total number of failed tests is <= 4, then the build/test job is marked a success. This is not ideal because the flush test may pass (it has passed 2-3 times out of the 15 times I ran the job today) and some other valid test may fail.
I am also wondering if we are testing this against latest released version of cb server, then the FTS is still dev-preview, right? In which case, should we make the FTS tests part of gerrit check? Should we just run only unit tests to pass the gerrit code review and then run integration tests as a later step by QE?