remove deps on data plugin#121955
Conversation
|
@elasticmachine merge upstream |
04c47cc to
72f9983
Compare
|
@elasticmachine merge upstream |
b846d09 to
c19eb90
Compare
c19eb90 to
c45f61c
Compare
|
@elasticmachine merge upstream |
15ae512 to
883cc99
Compare
|
@elasticmachine merge upstream |
|
buildkite, test this |
|
@elasticmachine merge upstream |
|
@elasticmachine merge upstream |
|
FYI: I merged |
|
@elasticmachine merge upstream |
|
Updating from |
|
@elasticmachine merge upstream |
|
buildkite, test this |
|
buildkite, test this |
💚 Build SucceededMetrics [docs]Module Count
Page load bundle
Unknown metric groupsESLint disabled line counts
Total ESLint disabled count
History
To update your PR or re-run it, just comment with: |
Summary
This pr highlights a blocker for removing
data_enhancedplugin in #122075To remove
data_enhanced, we will needdataplugin optionally depend onsecurityplugin.After fixing some circular dependency CI starts flaky failing on commit where
securitybecomes a dependency ofdataplugin: df7151aThe culprit is not clear yet.
Here you can see how I had 3 successful runs before adding
securityas dependency and then a failure after adding it in: df7151aSome observations:
status page should display the server status. Kibana reportsYellowstatus whenGreenis expected. After adding more logging I saw thattaskManageris causingYellowstatus.logs
logs
Reporting Functional Tests with Security enabled Security with reporting_user built-in role Dashboard: Download CSV file does not allow user that does not have reporting privileges. It fails on page load with timeout. logs. It is also often SAML / OIDC related tests that are likely to failerror: Setup lifecycle of "infra" plugin wasn't completed in 10sec. Consider disabling the plugin and re-start.I guess this just highlights that a bundle wasn't loaded on timeWhat we've tried:
securityplugin a dependency of a different OSS plugin (discover) - didn't reproduce the issue. So issue is reproduced ondatabut not ondiscoverUPDATE 1
In #123220 we've added debug logs for security and taskManager
We've noticed that, for example, in a very simple test:
Call to es takes around 30 seconds!!
The test fails because the token expires before it can continue.
logs
Update 2
We've tried adding a delay between es + kibana start and tests run: 2da369f
🟢 🟢 Tests started passing! 🟢 🟢
This means there is clearly some instablity in the beginning causing kibana / elasticsearch slow to respond and tests to fail