Move output dir logic to suite #29580
Conversation
clarkb7
added
changelog/no-changelog
team/windows-agent
qa/no-code-change
|
[Fast Unit Tests Report] On pipeline 46027273 (CI Visibility). The following jobs did not run any unit tests: Jobs:
If you modified Go files and expected unit tests to run in these jobs, please double check the job logs. If you think tests should have been executed reach out to #agent-devx-help |
Test changes on VMUse this command from test-infra-definitions to manually test this PR changes on a VM: inv create-vm --pipeline-id=46027273 --os-family=ubuntuNote: This applies to commit 124d8700 |
Regression DetectorRegression Detector ResultsRun ID: 39626e13-063b-4f72-a02a-17190d1dc670 Metrics dashboard Target profiles Baseline: 2abc3be Performance changes are noted in the perf column of each table:
No significant changes in experiment optimization goalsConfidence level: 90.00% There were no significant changes in experiment optimization goals at this confidence level and effect size tolerance.
|
| perf | experiment | goal | Δ mean % | Δ mean % CI | trials | links |
|---|---|---|---|---|---|---|
| ➖ | pycheck_lots_of_tags | % cpu utilization | +2.31 | [-0.18, +4.81] | 1 | Logs |
| ➖ | tcp_syslog_to_blackhole | ingress throughput | +0.71 | [+0.66, +0.76] | 1 | Logs |
| ➖ | otel_to_otel_logs | ingress throughput | +0.29 | [-0.52, +1.09] | 1 | Logs |
| ➖ | idle_all_features | memory utilization | +0.25 | [+0.15, +0.34] | 1 | Logs |
| ➖ | tcp_dd_logs_filter_exclude | ingress throughput | -0.00 | [-0.01, +0.01] | 1 | Logs |
| ➖ | uds_dogstatsd_to_api | ingress throughput | -0.02 | [-0.11, +0.07] | 1 | Logs |
| ➖ | idle | memory utilization | -0.19 | [-0.24, -0.15] | 1 | Logs |
| ➖ | file_tree | memory utilization | -0.47 | [-0.58, -0.36] | 1 | Logs |
| ➖ | uds_dogstatsd_to_api_cpu | % cpu utilization | -0.80 | [-1.53, -0.07] | 1 | Logs |
| ➖ | basic_py_check | % cpu utilization | -0.83 | [-3.59, +1.93] | 1 | Logs |
Bounds Checks
| perf | experiment | bounds_check_name | replicates_passed |
|---|---|---|---|
| ✅ | idle | memory_usage | 10/10 |
Explanation
A regression test is an A/B test of target performance in a repeatable rig, where "performance" is measured as "comparison variant minus baseline variant" for an optimization goal (e.g., ingress throughput). Due to intrinsic variability in measuring that goal, we can only estimate its mean value for each experiment; we report uncertainty in that value as a 90.00% confidence interval denoted "Δ mean % CI".
For each experiment, we decide whether a change in performance is a "regression" -- a change worth investigating further -- if all of the following criteria are true:
-
Its estimated |Δ mean %| ≥ 5.00%, indicating the change is big enough to merit a closer look.
-
Its 90.00% confidence interval "Δ mean % CI" does not contain zero, indicating that if our statistical model is accurate, there is at least a 90.00% chance there is a difference in performance between baseline and comparison variants.
-
Its configuration does not mark it "erratic".
| if os.Getenv("CI") == "" { | ||
| // Create a symlink to the latest run for user convenience | ||
| // TODO: Is there a standard "ci" vs "local" check? | ||
| // This code used to be in localProfile.GetOutputDir() | ||
| latestLink := filepath.Join(filepath.Dir(outputRoot), "latest") | ||
| // Remove the symlink if it already exists | ||
| if _, err := os.Lstat(latestLink); err == nil { | ||
| err = os.Remove(latestLink) | ||
| if err != nil { | ||
| return "", err | ||
| } | ||
| } | ||
| err = os.Symlink(outputRoot, latestLink) | ||
| if err != nil { | ||
| return "", err | ||
| } | ||
| } |
There was a problem hiding this comment.
💬 suggestion
What about passing the profile, and having the profile passing a path to the output root directory. Then each suite would create a directory inside the root directory, while the generic one is static.
There was a problem hiding this comment.
I'm not sure I'm understanding.
profile.GetOutputDir() already returns static output dir, e.g. ~/e2e-output. The suite calls GetRootOutputDir() to create a new subdirectory for itself.
There was a problem hiding this comment.
The rename to Create fixes this too
There was a problem hiding this comment.
What I meant in my message is that is profile that knows where you are running. You might have a generic profile GetLatestLink that cleans up symlinks only on localprofile and that does nothing on ci profile. You would call it here with runner.GetProfile().CleanupOldLinks() or a better name.
|
/merge |
|
🚂 MergeQueue: waiting for PR to be ready This merge request is not mergeable yet, because of pending checks/missing approvals. It will be added to the queue as soon as checks pass and/or get approvals. Use |
|
/merge -c |
|
This merge request was unqueued If you need support, contact us on Slack #devflow! |
|
/merge |
|
🚂 MergeQueue: waiting for PR to be ready This merge request is not mergeable yet, because of pending checks/missing approvals. It will be added to the queue as soon as checks pass and/or get approvals. Use |
|
🚂 MergeQueue: pull request added to the queue The median merge time in Use |
|
/merge |
|
🚂 MergeQueue: pull request added to the queue The median merge time in Use |
What does this PR do?
Moves E2E output dir logic from the profile to the suite, fixing an issue that created more directories than intended.
previous calls
new calls
Motivation
profile.GetOutputDir()was supposed to create a timestamped directory once and save it, but it was a value receiver method so the directory was never being saved, leading to a new timestamped directory being created on each call toGetTestOutputDir().Describe how to test/QA your changes
N/A, test change
can check that
new-e2e-windows-service-testonly creates 2 timestamped dirs (one for each suite), where previously it made many more.Possible Drawbacks / Trade-offs
Removes "tests" related things from runner/profile, but jobs/runs that execute multiple suites, or the same suite multiple times, will still create multiple timestamped directories. This does help avoid directory collisions for parameterized test suites, though ideally the test author would also make the test name unique by using
t.Run("<params>", ...).Moved symlink logic from
localProfile.GetOutputDir()to behind aos.Getenv("CI")condition, but would like a better/standardized way of determining "ci profile" vs "local profile" from outside the profile.Keeping
profile.GetOutputDir()method rather than only usingprofile.ParamStore()because local and CI profiles have different behavior for the default root dir$CI_PROJECT_DIR~/e2e-outputMoved the
runner.GetProfile()into the call itself, since that's what the suite does elsewhere.The usage in agent_client.go doesn't have access to the suite so it can't benefit from the root output dir caching done by the suite, but it does have access to the
e2e.Contextinterface (which is the suite under the hood). Maybe we can add output dir methods to thee2e.Contextinterface.Additional Notes
overall logic/path format,
<root>/<timestamp>/<test name>, is unchanged, but improvements can be made in follow ups, like maybe dropping the timestamp in the CI, or including part of the test name in the timestamp part as a hint when multiple suites are run and create multiple timestamp dirs.