Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Don't fail if benchmark performs better irrespective of threshold config #120

Open
wants to merge 1 commit into
base: master
Choose a base branch
from
Open
Changes from all commits
Commits
File filter

Filter by extension

Filter by extension

Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
Don't fail if benchmark performs better irrespective of threshold config
There might be situations if the changes drastically improves the
performance or the thresholds are configured with lower values it lead
to failure of the action whereas it is not.

i.e if the alert and failure threshold are configured as 0.1 and 0.2 and
the previous and current values are bigger, then the existing check
would consider it as failure

So added a check to ensure the failure is calculated based on the
performance and threshold rather than threshold alone
  • Loading branch information
madhan-mettle committed Apr 29, 2022
commit ca016f17ac58e0138acc2aa6d0e63bdf7845213f
14 changes: 11 additions & 3 deletions src/write.ts
Original file line number Diff line number Diff line change
@@ -85,6 +85,7 @@ interface Alert {
current: BenchmarkResult;
prev: BenchmarkResult;
ratio: number;
isPerformanceBetter: boolean;
}

function findAlerts(curSuite: Benchmark, prevSuite: Benchmark, threshold: number): Alert[] {
@@ -98,7 +99,9 @@ function findAlerts(curSuite: Benchmark, prevSuite: Benchmark, threshold: number
continue;
}

const ratio = biggerIsBetter(curSuite.tool)
const isBiggerBetter = biggerIsBetter(curSuite.tool);
const isPerformanceBetter = isBiggerBetter ? current.value > prev.value : current.value < prev.value;
const ratio = isBiggerBetter
? prev.value / current.value // e.g. current=100, prev=200
: current.value / prev.value; // e.g. current=200, prev=100

@@ -107,7 +110,7 @@ function findAlerts(curSuite: Benchmark, prevSuite: Benchmark, threshold: number
`Performance alert! Previous value was ${prev.value} and current value is ${current.value}.` +
` It is ${ratio}x worse than previous exceeding a ratio threshold ${threshold}`,
);
alerts.push({ current, prev, ratio });
alerts.push({ current, prev, ratio, isPerformanceBetter });
}
}

@@ -297,7 +300,12 @@ async function handleAlert(benchName: string, curSuite: Benchmark, prevSuite: Be
// Note: alertThreshold is smaller than failThreshold. It was checked in config.ts
const len = alerts.length;
const threshold = floatStr(failThreshold);
const failures = alerts.filter((a) => a.ratio > failThreshold);

//Only consider it as failure if the benchmark performed better irrespective of the alert threshold
// For ex if there is a massive performance improvement between commits
// or the alert threshold and failure threshold is lower the failure might be triggered.
// This check ensures that the check is failure only if the performance is not better
Comment on lines +304 to +307
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

This comment is difficult to read. It starts with "Only consider it as failure if the benchmark performed better…" Isn't the purpose of this change not to fail if the benchmark performs better?

const failures = alerts.filter((a) => !a.isPerformanceBetter).filter((a) => a.ratio > failThreshold);
if (failures.length > 0) {
core.debug('Mark this workflow as fail since one or more fatal alerts found');
if (failThreshold !== alertThreshold) {
25 changes: 25 additions & 0 deletions test/write.spec.ts
Original file line number Diff line number Diff line change
@@ -727,6 +727,31 @@ describe.each(['https://github.com', 'https://github.enterprise.corp'])('writeBe
},
error: undefined,
},
{
it: 'does not raise an alert when the benchmark is better irrespective of the alertThreshold',
config: { ...defaultCfg, alertThreshold: 0.01, failThreshold: 0.01 },
data: {
lastUpdate,
repoUrl,
entries: {
'Test benchmark': [
{
commit: commit('prev commit id'),
date: lastUpdate - 1000,
tool: 'go',
benches: [bench('bench_fib_10', 14)],
},
],
},
},
added: {
commit: commit('current commit id'),
date: lastUpdate,
tool: 'go',
benches: [bench('bench_fib_10', 11)], // Exceeds 1.0 alertthreshold
},
error: undefined,
},
];

for (const t of normalCases) {