Skip to content

[7.10] [Security Solution][Detections] Fix EQL cypress tests (#80440)#81211

Merged
rylnd merged 1 commit intoelastic:7.10from
rylnd:backport/7.10/pr-80440
Oct 20, 2020
Merged

[7.10] [Security Solution][Detections] Fix EQL cypress tests (#80440)#81211
rylnd merged 1 commit intoelastic:7.10from
rylnd:backport/7.10/pr-80440

Conversation

@rylnd
Copy link
Copy Markdown
Contributor

@rylnd rylnd commented Oct 20, 2020

Backports the following commits to 7.10:

* Unskip EQL tests

These _should_ be fixed with the latest ES on master, but let's see if
CI disagrees.

* Wait until alerts have populated on Rule Details

Occasionally our tests hit a scenario where the rule has executed (its
status is "succeeded"), but the generated alerts have not populated in
the same time frame. In this case the test fails oddly, saying that the
"alert count" element is not there when it is.

I attempted to improve the error message by using a .should() with a
callback, but that lead to even stranger behavior as the .should() would
fail once (expected), and then not be able to find the element a second
time. :(

So we instead focus on fixing the real problem, here: wait until alerts
populate (have a non-zero count) before performing the assertion.
Because the page will not update automatically, we can't rely on
cypress' retryability and must instead assert, click Refresh, and assert
again, much like we're doing while waiting for the rule to execute. And
like `waitForTheRuleToBeExecuted`, we're using a while loop that has no
guarantee of ever exiting :(

* More robust cypress assertions

* Uses should with a text matcher instead of using invoke('text')
* Use of not.equal between a string and an element may have been a false
  positive

* Perform cypress loops in a manner guaranteed to exit

We have a few tasks that require polling for some background work to be
completed. The basic form is: assert the byproduct, or refresh the page
and try again.

We were previously doing this with a while loop, which was not
guaranteed to ever complete, leading to cryptic failures if the process
ever hung.

Instead, this implements a safer polling mechanism with a definite
termination similar to the cypress-wait-until plugin.

* Update other specs that are asserting on alerts

* Do not automatically refresh the page
  * This is only necessary if we're not in the state we need. The
    `waitFor` helper functions automatically reload whatever needs to be
    reloaded, so we're delegating this task to them.
* Ensure we wait for alerts to be nonzero before our assertion
  * Otherwise we get some strange behavior around this field's
    availability; see previous commits

* Remove unused import

* Fix false positive in Rule Creation specs

Threat Match Rules introduced an additional query input, causing our
CUSTOM_QUERY_INPUT to be ambiguous.

However, instead of failing due to the ambiguity, the behavior of
cypress seems to be to pass! While I haven't yet tracked down the cause
of these false positives, disambiguating these selectors is the
immediate fix.

Co-authored-by: Kibana Machine <42973632+kibanamachine@users.noreply.github.com>
# Conflicts:
#	x-pack/plugins/security_solution/cypress/integration/alerts_detection_rules_eql.spec.ts
@rylnd rylnd added the backport This PR is a backport of another PR label Oct 20, 2020
@kibanamachine
Copy link
Copy Markdown
Contributor

💚 Build Succeeded

Metrics [docs]

✅ unchanged

To update your PR or re-run it, just comment with:
@elasticmachine merge upstream

@rylnd rylnd merged commit 3fc1f8c into elastic:7.10 Oct 20, 2020
@rylnd rylnd deleted the backport/7.10/pr-80440 branch October 20, 2020 19:07
@rylnd
Copy link
Copy Markdown
Contributor Author

rylnd commented Oct 20, 2020

Mentioned there, but: this also backports #79287 to 7.10.

spalger pushed a commit that referenced this pull request Oct 20, 2020
* Unskip EQL tests

These _should_ be fixed with the latest ES on master, but let's see if
CI disagrees.

* Wait until alerts have populated on Rule Details

Occasionally our tests hit a scenario where the rule has executed (its
status is "succeeded"), but the generated alerts have not populated in
the same time frame. In this case the test fails oddly, saying that the
"alert count" element is not there when it is.

I attempted to improve the error message by using a .should() with a
callback, but that lead to even stranger behavior as the .should() would
fail once (expected), and then not be able to find the element a second
time. :(

So we instead focus on fixing the real problem, here: wait until alerts
populate (have a non-zero count) before performing the assertion.
Because the page will not update automatically, we can't rely on
cypress' retryability and must instead assert, click Refresh, and assert
again, much like we're doing while waiting for the rule to execute. And
like `waitForTheRuleToBeExecuted`, we're using a while loop that has no
guarantee of ever exiting :(

* More robust cypress assertions

* Uses should with a text matcher instead of using invoke('text')
* Use of not.equal between a string and an element may have been a false
  positive

* Perform cypress loops in a manner guaranteed to exit

We have a few tasks that require polling for some background work to be
completed. The basic form is: assert the byproduct, or refresh the page
and try again.

We were previously doing this with a while loop, which was not
guaranteed to ever complete, leading to cryptic failures if the process
ever hung.

Instead, this implements a safer polling mechanism with a definite
termination similar to the cypress-wait-until plugin.

* Update other specs that are asserting on alerts

* Do not automatically refresh the page
  * This is only necessary if we're not in the state we need. The
    `waitFor` helper functions automatically reload whatever needs to be
    reloaded, so we're delegating this task to them.
* Ensure we wait for alerts to be nonzero before our assertion
  * Otherwise we get some strange behavior around this field's
    availability; see previous commits

* Remove unused import

* Fix false positive in Rule Creation specs

Threat Match Rules introduced an additional query input, causing our
CUSTOM_QUERY_INPUT to be ambiguous.

However, instead of failing due to the ambiguity, the behavior of
cypress seems to be to pass! While I haven't yet tracked down the cause
of these false positives, disambiguating these selectors is the
immediate fix.

Co-authored-by: Kibana Machine <42973632+kibanamachine@users.noreply.github.com>
# Conflicts:
#	x-pack/plugins/security_solution/cypress/integration/alerts_detection_rules_eql.spec.ts
(cherry picked from commit 3fc1f8c)
@MindyRS MindyRS added the Team: SecuritySolution Security Solutions Team working on SIEM, Endpoint, Timeline, Resolver, etc. label Oct 27, 2020
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

backport This PR is a backport of another PR Team: SecuritySolution Security Solutions Team working on SIEM, Endpoint, Timeline, Resolver, etc.

Projects

None yet

Development

Successfully merging this pull request may close these issues.

3 participants