e2e tests reliability issues

[joeyorlando]

There are a few UI bugs which currently are contributing to the e2e tests not being reliable on CI. Below is a list of issues I have seen (so far):

• When creating a schedule via the UI, you sometimes get redirected to the schedule page and everything is grayed out #2423
• Schedule race condition - TypeError: undefined is not an object (evaluating 'dayOptions.forEach') #2802
• TypeError: Cannot read properties of undefined (reading 'forEach') #2820
• debug e2e test HTTP 50x responses that occasionally pop up #2236
• now with running things locally via helm + k8s we occasionally see the mariadb container crash, which in turn causes HTTP 50x from the grafana and oncall-engine containers. Solving this root issue will fix these HTTP 50xs:

  Events:                                                                                                                                                                                                                                                                    
  │   Type     Reason            Age                   From               Message                                                                                                                                                                                              
  │   ----     ------            ----                  ----               -------                                                                                                                                                                                                                                                        
  │   Normal   Created           18m                   kubelet            Created container mariadb
  │   Normal   Started           18m                   kubelet            Started container mariadb                                                                                                                                                                            
  │
  │   Warning  Unhealthy         7m22s (x11 over 15m)  kubelet            Liveness probe failed: command "/bin/bash -ec password_aux=\"${MARIADB_ROOT_PASSWORD:-}\"\nif [[ -f \"${MARIADB_ROOT_PASSWORD_FILE:-}\" ]]; then\n    password_aux=$(cat \"$MARIADB_ROOT_PASSWORD_FI ││ LE\")\nfi\nmysqladmin status -uroot -p\"${password_aux}\"\n" timed out                                                                                                                                                                                                     
  │   Warning  Unhealthy         2m42s (x18 over 15m)  kubelet            Readiness probe failed: command "/bin/bash -ec password_aux=\"${MARIADB_ROOT_PASSWORD:-}\"\nif [[ -f \"${MARIADB_ROOT_PASSWORD_FILE:-}\" ]]; then\n    password_aux=$(cat \"$MARIADB_ROOT_PASSWORD_F ││ ILE\")\nfi\nmysqladmin status -uroot -p\"${password_aux}\"\n" timed out
  

• Occasionally when creating a web schedule, the drawer does not close and redirect the user to the schedule detail page #2461
• Performance Issue in GSelect component when searching #1628 - this one is preventing us from being able to parallelize the tests due to the # of API requests that the UI makes + the very limited amount of CPU resources that the GitHub Actions CI host provides. Parallelization will be important once we start to amass more and more e2e tests.
• race condition when fetching schedule quality, just after creating an oncall-shift, leads to a "Bad" quality schedule ("Schedule is Empty") #1968
• When adding a schedule rotation, via the UI, the rotation layer sometimes disappears randomly #1655
• I have seen a handful of cases where the tests will fail globalSetup.ts, mostly related to a failing assertion when configuring the plugin. We should increase the timeout for the "plugin configured" assertion from the default of 5s to ~25s. This is because it can sometimes take a bit longer for the backend sync to finish (closed in re-enable e2e UI tests on CI #1961)

  // wait for the "Connected to OnCall" message to know that everything is properly configured
  await expect(page.getByTestId('status-message-block')).toHaveText(/Connected to OnCall.*/);