Sanitize principal names in AWS STS role session names by yushesp · Pull Request #3525 · apache/polaris

yushesp · 2026-01-24T17:11:32Z

Principal names containing invalid characters (spaces, parentheses, etc.) were causing AWS STS AssumeRole requests to fail with validation errors. AWS STS role session names must match the pattern [\w+=,.@-]*.

This change:

Adds AwsRoleSessionNameSanitizer utility class to sanitize strings for use as AWS STS role session names
Replaces invalid characters with underscores and truncates to 64 characters (AWS maximum)
Updates AwsCredentialsStorageIntegration to sanitize principal names when INCLUDE_PRINCIPAL_NAME_IN_SUBSCOPED_CREDENTIAL is enabled
Adds tests to verify sanitization behavior and AWS pattern compliance

Fixes issue where principal names like Joe (local) would produce invalid role session names like Polaris-Joe (local) and cause AssumeRole to fail. Now sanitized to polaris-Joe__local_.

Checklist

🛡️ Don't disclose security issues! (contact security@apache.org)
🔗 Clearly explained why the changes are needed, or linked related issues: Fixes #
🧪 Added/updated tests with good coverage, or manually tested (and explained how)
💡 Added comments for complex logic
🧾 Updated CHANGELOG.md (if needed)
📚 Updated documentation in site/content/in-dev/unreleased (if needed)

Principal names containing invalid characters (spaces, parentheses, etc.) were causing AWS STS AssumeRole requests to fail with validation errors. AWS STS role session names must match the pattern [\w+=,.@-]*. This change: - Adds AwsRoleSessionNameSanitizer utility class to sanitize strings for use as AWS STS role session names - Replaces invalid characters with underscores and truncates to 64 characters (AWS maximum) - Updates AwsCredentialsStorageIntegration to sanitize principal names when INCLUDE_PRINCIPAL_NAME_IN_SUBSCOPED_CREDENTIAL is enabled - Adds tests to verify sanitization behavior and AWS pattern compliance Fixes issue where principal names like "Joe (local)" would produce invalid role session names like "polaris-Joe (local)" and cause AssumeRole to fail. Now sanitized to "polaris-Joe__local_". Co-authored-by: carc-prathyush-shankar <prathyush.shankar@carbonarc.co>

snazy

Thanks for the fix @yushesp !

dimas-b · 2026-01-26T17:35:36Z

...core/src/main/java/org/apache/polaris/core/storage/aws/AwsCredentialsStorageIntegration.java

    String roleSessionName =
        includePrincipalNameInSubscopedCredential
-            ? "polaris-" + polarisPrincipal.getName()
+            ? AwsRoleSessionNameSanitizer.sanitize("polaris-" + polarisPrincipal.getName())


The problem this change fixes is specific to AWS, AFAIK, but the benefits of including the exact principal name may be applicable to other S3 systems, where the session name is less restricted. Cf. #3224

@tokoko : WDYT? Is having exact principal names critical for your use cases?

@yushesp : In your use cases, do you actually need some Principal info in the session name, or could you exclude it via the INCLUDE_PRINCIPAL_NAME_IN_SUBSCOPED_CREDENTIAL flag?

exact names are preferable of course, but in case the name causes a fail, sanitizing it makes sense. we're effectively kind of already doing the same by capping the length of the role name. The principal name included could be different for really really long principal names.

for our use case, we're using both s3 and minio, so erring on the side of s3 restrictions sounds good. for the general case, assuming s3 restrictions makes sense for me as well, otherwise we'd have to either 1) try a call w/o sanitization and fall back on sanitization or 2) try coding restrictions for each system. Neither seems like a good choice.

@dimas-b We do need to inject the principal name into the session name. We enabled the feature flag and ran into an error because my principal name had a special character.

dimas-b

Let's wait a couple of more days in case other people want to review (it's a Polaris Community Sprint day today, so some people may be unavailable).

yushesp · 2026-02-04T00:56:32Z

@dimas-b is this PR good to be merged?

dimas-b · 2026-02-04T01:15:29Z

Thanks for the ping, @yushesp 😉

Principal names containing invalid characters (spaces, parentheses, etc.) were causing AWS STS AssumeRole requests to fail with validation errors. AWS STS role session names must match the pattern [\w+=,.@-]*. This change: - Adds AwsRoleSessionNameSanitizer utility class to sanitize strings for use as AWS STS role session names - Replaces invalid characters with underscores and truncates to 64 characters (AWS maximum) - Updates AwsCredentialsStorageIntegration to sanitize principal names when INCLUDE_PRINCIPAL_NAME_IN_SUBSCOPED_CREDENTIAL is enabled - Adds tests to verify sanitization behavior and AWS pattern compliance Fixes issue where principal names like "Joe (local)" would produce invalid role session names like "polaris-Joe (local)" and cause AssumeRole to fail. Now sanitized to "polaris-Joe__local_". Co-authored-by: carc-prathyush-shankar <prathyush.shankar@carbonarc.co>

* Releasey: adjust workflwo for Apache org level secrets (apache#3647) See reference [INFRA-27430](https://issues.apache.org/jira/browse/INFRA-27430), requiring us to use `DOCKERHUB_USER` + `DOCKERHUB_TOKEN` instead of `DOCKERHUB_USERNAME` + `DOCKERHUB_TOKEN`. * Guides: fix setup scripts to yield correct exit code (apache#3612) The statements in the shell scripts for the setup services are often concatenated using `;`, which means that a previous' command exit code is _not_ propagated and the service, although it failed, is determined to be successful. This change updates those scripts to use `&&` for the statement concatenation. "Final" setup services (aka "polaris-setup") now have a final `sleep 120`. This is due to the behavior of `docker compose up --detach --wait`, which considers _any_ service (without dependants) that exits with exit code 0 as a failure, leading to that docker-compose command yielding an error code. That would break the guides testing code (apache#3553). That `sleep 120` in "polaris-setup" services does **not** cause a delay of the compose starting up - it is purely a "hack around" Docker Compose not having a notion of "setup services". To avoid merge conflicts, this change also: * updates affected `curl` invocations (as apache#3610) * removes superfluous `restart: "no"` * Add PolarisEventType int codes and remove unused before/after commit view/table values. (apache#3608) This is a continuation of apache#3418 where we agreed we should remove associate enums values from Enum definition. The code also adds a constructor to help not having to rely on ordinals() to to have enum codes. I incremented enum code assignments by 100 based on categories. I am happy to take feedback here. Does not change logic and removes only enums that are no longer used so behavior does not change. * docs: Add quick guide for downstream builds (apache#3601) * docs: Add quick guide for downstream builds * IntegrationTestsHelper: fix extract & merge logic (apache#3650) Both methods were flawed: * `extractFromAnnotatedElements` wasn't properly delegating to the class if the method isn't annotated * `mergeFromAnnotatedElements` wasn't properly prioritizing method properties over class properties. * Community page: Yong Zheng - PPMC Member (apache#3653) * chore(deps): update actions/checkout digest to de0fac2 (apache#3652) * Use `quarkus.package.jar.type` (apache#3644) Switch to `quarkus.package.jar.type` instead of the old `quarkus.package.type` build property as suggested by Quarkus build warning: ``` 2026-02-03T00:26:05.672955189Z WorkerExecutor Queue WARN Configuration property 'quarkus.package.type' has been deprecated and replaced by: [quarkus.package.jar.enabled, quarkus.package.jar.type, quarkus.native.enabled, quarkus.native.sources-only] ``` * Site: Add blog post for Floe Polaris Integration (apache#3645) * "Stale" job: restrict executions and adjust issue permissions (apache#3636) * Add copyright on website (apache#3659) * Sanitize principal names in AWS STS role session names (apache#3525) Principal names containing invalid characters (spaces, parentheses, etc.) were causing AWS STS AssumeRole requests to fail with validation errors. AWS STS role session names must match the pattern [\w+=,.@-]*. This change: - Adds AwsRoleSessionNameSanitizer utility class to sanitize strings for use as AWS STS role session names - Replaces invalid characters with underscores and truncates to 64 characters (AWS maximum) - Updates AwsCredentialsStorageIntegration to sanitize principal names when INCLUDE_PRINCIPAL_NAME_IN_SUBSCOPED_CREDENTIAL is enabled - Adds tests to verify sanitization behavior and AWS pattern compliance Fixes issue where principal names like "Joe (local)" would produce invalid role session names like "polaris-Joe (local)" and cause AssumeRole to fail. Now sanitized to "polaris-Joe__local_". Co-authored-by: carc-prathyush-shankar <prathyush.shankar@carbonarc.co> * Site: Adds the copyright message to all site pages (apache#3661) * Site: Change Security link to local security reporting page (apache#3662) * fix(deps): update dependency org.mongodb:mongodb-driver-sync to v5.6.3 (apache#3654) * Releasey: use Apache org Nexus credentials (apache#3651) * Releasey: use the correct SVN credentials (apache#3648) * CI: Prerequisite PR for apache#3625 (apache#3646) This change is only needed to update the required-checks to be able to eventually merge apache#3625. * CI: all-in-one workflow (apache#3625) This change moves all CI jobs into a single workflow. A single workflow comes with a couple advantages: * all jobs are visible on one page * option to re-run all failed jobs at once The refactoring also simplifies the `.asf.yaml` file by referencing a single required check that can only succeeds if dependent jobs were successful. The actual CI jobs are in the `ci.yml` file, which is only triggered via a `workflow_call` event. There are two workflows that call `ci.yml`: * `ci-main.yml` for `main` and `release/*` branches, with a concurrency group that does not cancel already running workflows * `ci-pr.yml` for PRs with a concurrency group that cancels previous CI runs The names of these two calling workflows include information that enrich the workflow view, and the workflows for main, release branches and PRs are grouped separartely in the GH Actions page for the repository. In other words, the reference name (for main + release branches) or the PR number and title are shown in the workflow runs list on the GH Actions page. * Fix CI required checks (apache#3666) * Add Sung as committer (apache#3665) * Fix `CatalogFederationIntegrationTest.testFederatedCatalogWithCredentialVending()` for AWSSDK update (apache#3664) Recent AWSSDK versions introduce `software.amazon.awssdk.services.s3.model.AccessDeniedException`, hence the assertion on `S3Exception` fails. * CI/main: re-add commit message to `run_name` (apache#3668) The default `run_name` value is, in case of `push` events, the commit message. This change re-adds the commit message. * Add CI workflow to test against Iceberg unreleased versions (apache#3630) * Update dependency software.amazon.awssdk:bom to v2.41.21 (apache#3639) * Update actions/checkout digest to de0fac2 (apache#3671) * Nit: Fix wrong `Nullable` import (apache#3672) * Explicitly set build-time property `quarkus.datasource.db-kind` (apache#3674) The property is set for the Polaris admin tool, but not for Polaris server. This causes a startup error with Quarkus 3.31. Error message: ``` ERROR: Failed to start application java.lang.RuntimeException: Failed to start quarkus at io.quarkus.runner.ApplicationImpl.doStart(Unknown Source) at io.quarkus.runtime.Application.start(Application.java:116) at io.quarkus.runtime.ApplicationLifecycleManager.run(ApplicationLifecycleManager.java:119) at io.quarkus.runtime.Quarkus.run(Quarkus.java:79) at io.quarkus.runtime.Quarkus.run(Quarkus.java:50) at io.quarkus.runtime.Quarkus.run(Quarkus.java:143) at io.quarkus.runner.GeneratedMain.main(Unknown Source) at io.quarkus.bootstrap.runner.QuarkusEntryPoint.doRun(QuarkusEntryPoint.java:86) at io.quarkus.bootstrap.runner.QuarkusEntryPoint.main(QuarkusEntryPoint.java:37) Caused by: java.lang.IllegalStateException: Build time property cannot be changed at runtime: - quarkus.datasource.db-kind is set to 'postgresql' but it is build time fixed to 'null'. Did you change the property quarkus.datasource.db-kind after building the application? at io.quarkus.runtime.configuration.ConfigRecorder.handleConfigChange(ConfigRecorder.java:72) at io.quarkus.runner.recorded.ConfigGenerationBuildStep$checkForBuildTimeConfigChange1532146938.deploy_6(Unknown Source) at io.quarkus.runner.recorded.ConfigGenerationBuildStep$checkForBuildTimeConfigChange1532146938.deploy(Unknown Source) ... 9 more ``` * Last merged commit 2651e06 --------- Co-authored-by: Innocent Djiofack <djiofack007@gmail.com> Co-authored-by: Dmitri Bourlatchkov <dmitri.bourlatchkov@gmail.com> Co-authored-by: Alexandre Dutra <adutra@apache.org> Co-authored-by: Mend Renovate <bot@renovateapp.com> Co-authored-by: Neelesh Salian <nssalian@users.noreply.github.com> Co-authored-by: JB Onofré <jbonofre@apache.org> Co-authored-by: Prathyush Shankar <prathyush2018@gmail.com> Co-authored-by: carc-prathyush-shankar <prathyush.shankar@carbonarc.co> Co-authored-by: Russell Spitzer <russell.spitzer@GMAIL.COM>

github-project-automation bot added this to Basic Kanban Board Jan 24, 2026

github-project-automation bot moved this to PRs In Progress in Basic Kanban Board Jan 24, 2026

yushesp force-pushed the sanitize-role-session-name branch from 1af1fef to 867a27f Compare January 24, 2026 17:14

snazy approved these changes Jan 26, 2026

View reviewed changes

github-project-automation bot moved this from PRs In Progress to Ready to merge in Basic Kanban Board Jan 26, 2026

dimas-b reviewed Jan 26, 2026

View reviewed changes

adutra approved these changes Jan 27, 2026

View reviewed changes

dimas-b approved these changes Jan 27, 2026

View reviewed changes

snazy added this to the 1.4.0 milestone Jan 31, 2026

adnanhemani approved these changes Feb 2, 2026

View reviewed changes

dimas-b merged commit e37219d into apache:main Feb 4, 2026
15 checks passed

github-project-automation bot moved this from Ready to merge to Done in Basic Kanban Board Feb 4, 2026

yushesp deleted the sanitize-role-session-name branch February 4, 2026 01:18

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Sanitize principal names in AWS STS role session names#3525

Sanitize principal names in AWS STS role session names#3525
dimas-b merged 1 commit intoapache:mainfrom
yushesp:sanitize-role-session-name

yushesp commented Jan 24, 2026

Uh oh!

snazy left a comment

Uh oh!

dimas-b Jan 26, 2026

Uh oh!

tokoko Jan 26, 2026

Uh oh!

tokoko Jan 26, 2026

Uh oh!

yushesp Jan 27, 2026

Uh oh!

dimas-b left a comment

Uh oh!

yushesp commented Feb 4, 2026

Uh oh!

Uh oh!

dimas-b commented Feb 4, 2026

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

6 participants

Conversation

yushesp commented Jan 24, 2026

Checklist

Uh oh!

snazy left a comment

Choose a reason for hiding this comment

Uh oh!

dimas-b Jan 26, 2026

Choose a reason for hiding this comment

Uh oh!

tokoko Jan 26, 2026

Choose a reason for hiding this comment

Uh oh!

tokoko Jan 26, 2026

Choose a reason for hiding this comment

Uh oh!

yushesp Jan 27, 2026

Choose a reason for hiding this comment

Uh oh!

dimas-b left a comment

Choose a reason for hiding this comment

Uh oh!

yushesp commented Feb 4, 2026

Uh oh!

Uh oh!

dimas-b commented Feb 4, 2026

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

6 participants