Only explain sampled Postgres statements that are a single SQL statement by azhou-datadog · Pull Request #24839 · DataDog/integrations-core

azhou-datadog · 2026-08-12T13:48:30Z

What does this PR do?

The PREPARE that explains a parameterized query now goes out over the extended query protocol, where the server refuses a multi-command string outright instead of executing every statement in it.

A pipeline is the only route to that protocol here. psycopg leaves the simple protocol only for a query carrying parameters, one requesting binary results — which the pool's ClientCursor refuses — or one inside a pipeline.
It fails closed. Pipeline mode needs libpq 14+; without it the query goes unexplained rather than falling back to the simple protocol, and the check warns once at init. Agent 7 latest links libpq 18.3, so that branch should never be reached in a shipped Agent.
Verified on Postgres 16. A planted SELECT 1; CREATE TEMP TABLE …; --$1 is refused with cannot insert multiple commands into a prepared statement, and the table never appears. That is an integration test; the suite is 773 passed, 4 skipped.

Motivation

Text sampled from pg_stat_activity.query is interpolated into PREPARE dd_<signature> AS <statement> and run as the monitoring user over the simple query protocol, which executes every ;-separated statement. Any user who can query a monitored database can plant one and have the agent run it with the monitoring role's privileges. The feature is on by default.

https://datadoghq.atlassian.net/browse/VULN-92306

Review checklist (to be filled by reviewers)

Feature or bugfix MUST have appropriate tests (unit, integration, e2e)
Add qa/required if this PR needs QA validation, or qa/skip-qa if it does not. Exactly one of the two is required.
If you need to backport this PR to another branch, you can add the backport/<branch-name> label to the PR and it will automatically open a backport PR once this one is merged

Sampled statement text is read out of pg_stat_activity and interpolated into a PREPARE that the agent executes as the monitoring user, using the simple query protocol. Anyone able to run a query on a monitored database could therefore plant statement separators in the text and have the agent run them with the monitoring role's privileges. Reject any sampled statement that is not unambiguously one top-level statement, before either explain path touches the database, and again at the PREPARE sink itself. Text that cannot be tokenized with certainty (unterminated literals or comments, a backslash in a plain string whose meaning depends on standard_conforming_strings) is refused rather than guessed at. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>

Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>

cit-pr-commenter-54b7da · 2026-08-12T13:51:35Z

evalya-impact-summary

evalya impact analysis

Impact analysis: 0 selected, 0 skipped (of 0 test tasks)
Publish tasks:   0 (always emitted)
Diff (4 files):
  postgres/changelog.d/24839.security
  postgres/datadog_checks/postgres/explain_parameterized_queries.py
  postgres/datadog_checks/postgres/util.py
  postgres/tests/test_explain_parameterized_queries.py

Debug a specific task: evalya plan impact --path <path> --task <task>

Learn more about CI impact filtering

datadog-official · 2026-08-12T13:53:15Z

🎉 All green!

🧪 All tests passed
❄️ No new flaky tests detected

🚧 1 test that failed was ignored due to quarantine

🎯 Code Coverage (details)
• Patch Coverage: 93.88%
• Overall Coverage: 94.17% (+5.72%)

_{This comment will be updated automatically if new data arrives.

🔗 Commit SHA: c21a78b | Docs | Datadog PR Page | Give us feedback!}

psycopg only leaves the simple query protocol, which executes every ;-separated statement in a string, when a query carries parameters, asks for binary results or runs in a pipeline. The pool's client-side cursors reject binary results, so a pipeline is the only route: wrap the PREPARE in one, and the server itself refuses a multi-command string. Guarded by a libpq 14+ capability check, since pipeline mode needs it. This puts the server rather than the single-statement scan in charge of the boundary. The scan stays: it reports a clean error code instead of a syntax error, and it is the only guard where libpq is older than 14. Also trims the comments and docstrings this branch added, and drops a stale example from the pre-existing datatype_mismatch comment: a multi-statement string sent through datadog.explain_statement raises InvalidCursorDefinition, not a datatype mismatch. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>

The pipeline makes the server reject sampled text holding more than one statement, so the hand-written tokenizer was a second, weaker implementation of the same rule, and the only thing it still covered was a build whose libpq is too old for pipeline mode. Fail closed there instead: without pipeline support the query goes unexplained, with one warning at init, rather than falling back to the simple query protocol. Explaining a parameterized query is worth less than the guarantee that sampled text never runs. Also reverts the test_failed_explain_handling expectation. That statement has no $1, so it takes the datadog.explain_statement path, which rejects a multi-query plan on its own and still reports InvalidCursorDefinition. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>

chatgpt-codex-connector

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: 3719462adc

ℹ️ About Codex in GitHub

Codex has been enabled to automatically review pull requests in this repo. Reviews are triggered when you

Open a pull request for review
Mark a draft as ready
Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

When you sign up for Codex through ChatGPT, Codex can also answer questions or update the PR, like "@codex address that feedback".

test_execute_prepare_uses_a_pipeline asserted the pipeline branch while taking _can_use_pipeline from whatever libpq the machine links, so it would fail rather than skip on a build older than 14. Set it explicitly, as the fail-closed test already does. The integration test had the same dependence. Its injection assertion holds either way -- fail closed sends nothing -- so only the error-class assertion is now conditional. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>

dd-octo-sts · 2026-08-14T16:00:23Z

Validation Report

All 21 validations passed.

Show details

Validation	Description	Status
`agent-reqs`	Verify check versions match the Agent requirements file	✅
`ci`	Validate CI configuration and code coverage settings	✅
`codeowners`	Validate every integration has a CODEOWNERS entry	✅
`config`	Validate default configuration files against spec.yaml	✅
`dep`	Verify dependency pins are consistent and Agent-compatible	✅
`http`	Validate integrations use the HTTP wrapper correctly	✅
`imports`	Validate check imports do not use deprecated modules	✅
`integration-style`	Validate check code style conventions	✅
`jmx-metrics`	Validate JMX metrics definition files and config	✅
`labeler`	Validate PR labeler config matches integration directories	✅
`legacy-signature`	Validate no integration uses the legacy Agent check signature	✅
`license-headers`	Validate Python files have proper license headers	✅
`licenses`	Validate third-party license attribution list	✅
`metadata`	Validate metadata.csv metric definitions	✅
`models`	Validate configuration data models match spec.yaml	✅
`openmetrics`	Validate OpenMetrics integrations disable the metric limit	✅
`package`	Validate Python package metadata and naming	✅
`qa-label`	Validate the pull request declares whether it needs QA for the next Agent release	✅
`readmes`	Validate README files have required sections	✅
`saved-views`	Validate saved view JSON file structure and fields	✅
`version`	Validate version consistency between package and changelog	✅

View full run

temporal-github-worker-1 Bot added agent/review-requested ecosystems/review-requested product/review-requested labels Aug 12, 2026

dd-octo-sts Bot added the integration/postgres label Aug 12, 2026

azhou-datadog added the qa/required QA is required for this PR and will generate a QA card label Aug 12, 2026

Add changelog entry

e86499d

Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>

azhou-datadog mentioned this pull request Aug 12, 2026

Only explain sampled Postgres statements that are a single SQL statement #24838

Closed

3 tasks

azhou-datadog and others added 2 commits August 12, 2026 10:46

azhou-datadog marked this pull request as ready for review August 12, 2026 17:03

azhou-datadog requested review from a team as code owners August 12, 2026 17:04

dd-octo-sts Bot added team/agent-integrations team/database-monitoring-agent labels Aug 12, 2026

chatgpt-codex-connector Bot reviewed Aug 12, 2026

View reviewed changes

Comment thread postgres/tests/test_explain_parameterized_queries.py

azhou-datadog and others added 2 commits August 12, 2026 14:06

remove comments mentioning vuln ticket

c21a78b

eric-weaver approved these changes Aug 14, 2026

View reviewed changes

Navigation Menu

Uh oh!

Only explain sampled Postgres statements that are a single SQL statement - #24839

Only explain sampled Postgres statements that are a single SQL statement#24839
azhou-datadog wants to merge 6 commits into
masterfrom
allen.zhou/vuln-92306-clean

azhou-datadog commented Aug 12, 2026 •

edited

Loading

Uh oh!

cit-pr-commenter-54b7da Bot commented Aug 12, 2026 •

edited

Loading

Uh oh!

datadog-official Bot commented Aug 12, 2026 •

edited by datadog-datadog-prod-us1 Bot

Loading

Uh oh!

chatgpt-codex-connector Bot left a comment

Uh oh!

Uh oh!

dd-octo-sts Bot commented Aug 14, 2026

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

Uh oh!

Conversation

azhou-datadog commented Aug 12, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

What does this PR do?

Motivation

Review checklist (to be filled by reviewers)

Uh oh!

cit-pr-commenter-54b7da Bot commented Aug 12, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

evalya-impact-summary

Uh oh!

datadog-official Bot commented Aug 12, 2026 • edited by datadog-datadog-prod-us1 Bot Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

chatgpt-codex-connector Bot left a comment

Choose a reason for hiding this comment

💡 Codex Review

Uh oh!

Uh oh!

dd-octo-sts Bot commented Aug 14, 2026

Validation Report

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

azhou-datadog commented Aug 12, 2026 •

edited

Loading

cit-pr-commenter-54b7da Bot commented Aug 12, 2026 •

edited

Loading

datadog-official Bot commented Aug 12, 2026 •

edited by datadog-datadog-prod-us1 Bot

Loading