kubeflow/pipelines: technical due diligence, read from its repository

The median source file here is 106 lines. 25 files exceed 1,060 lines, the largest being 4,431. Size is not a defect on its own — some problems genuinely live in one place — but these are where merge conflicts, review fatigue and single-owner knowledge concentrate, and they are the first thing that slows an inheriting engineer down.

The suite is configured to retry failing tests (test/frontend-integration-test/dependency-smoke.test.mjs). A retry turns a flaky test green without saying why it failed, so a green build means the tests passed eventually. Ask how many retries a typical run needs, and which tests need them.

Deleted lines per line added rose from 0.40 to 0.82 between the first and second halves of this history. Rework is normal and often healthy, but a rising ratio means an increasing share of the team's time is spent replacing code that already exists rather than building what does not. Whether that is deliberate repayment or churn is a question for the team.

15 files were marked deprecated over two years ago and still are, the oldest since Oct 2019. A deprecation is a promise to remove. Ask what still depends on these, and whether removing them is scheduled.

The newest release, 2.18.0, is not on the branch this survey read; the newest release on it is 2.3.0. The shipped product may differ from what this report describes. Survey the release branch too.

12 HTTP calls use a client with no timeout (Python `requests` without `timeout=`, or Go's default client). When a dependency slows down, each one holds a worker until the connection gives up, and enough of them take the service down with the dependency.

1 places in product code build SQL by formatting or joining strings rather than passing parameters. In each, what the code shows going in is a table or column name, a constant, or something the line does not show, not a request's value. Ask whether any value reaches these strings, rather than a parameter.

27 workload manifests set no CPU or memory limits. One pod that leaks memory or spins can starve everything else on its node, and the scheduler cannot plan capacity for what it cannot see.

Third-party code is committed into this repository without any statement of the terms it arrived under. Unknown terms are harder to deal with than inconvenient ones: an awkward licence can be complied with, an absent one cannot. Each of these needs its origin established before the code changes hands.

The newest was published on 2026-10-29. A project that publishes advisories is telling its users about its fixes, and the count and pace say how much security work it carries. Whether 2 of them are fixed in the code you ship could not be told from this history: its version is only known to be at least 2.3.0. Answer lives in: the release you run, set against each fixed version. GHSA-w95g-prjh-32jh (high), GHSA-m6q4-2p22-7x6f (moderate), GHSA-r98w-fgcr-qc52 (moderate) and 14 more are fixed in the surveyed commit; which release carries them cannot be told from this history, which does not place the release users run. Answer lives in: the version you run. The record is incomplete: NVD: 60 records name pipelines with nothing tying them to this repository (CVE-2023-37264, CVE-2024-38363, CVE-2024-41950, …).

Nothing in the service initialises an error reporter (Sentry, Rollbar, Bugsnag, Honeybadger or an APM agent). An exception in production reaches a person only if someone is reading the logs, which usually means a customer reports it first.

4 deployments set `replicas: 1`, and nothing (an autoscaler, a disruption budget) adds more. Every deploy and every node drain is downtime, and the availability the buyer inherits has a ceiling.