How to Configure ArgoCD SSO with AWS IAM Identity Center Using Dex SAML

·1 views

A technical guide details how to integrate ArgoCD single sign-on with AWS IAM Identity Center using ArgoCD's built-in Dex as a SAML service provider. The setup works with both IAM Identity Center's native directory and external identity providers such as Google, requiring no client secrets since trust is established via certificate. Key configuration steps include creating a custom SAML 2.0 application in Identity Center, mapping user attributes, and updating ArgoCD's config map with the SSO URL and certificate data. The guide highlights several non-obvious pitfalls, including the need to set the Subject NameID format to 'persistent' and ensuring the SAML audience matches the entityIssuer to avoid cryptic errors. Once SSO is verified, the built-in local admin account can be disabled to enforce identity-provider-only access.

Read the full story at DEV Community

This is an AI-generated summary. ShortSingh links to the original source for the complete article.

Discussion (0)

Experiment Shows 'Persist-Then-Ack' Webhook Pattern Prevents Silent Event Loss

A software developer ran a controlled experiment to test whether webhook handler design affects silent data loss during process crashes. Two variants were compared across 50 trials each: one that acknowledges delivery before saving to the database, and one that saves first then acknowledges. The ack-then-persist approach lost events in all 50 trials, as the process was killed after sending a 200 response but before the database write completed, leaving the provider with no reason to retry. The persist-then-ack variant lost zero events, since a committed write before acknowledgment ensures retries can recover any unacknowledged delivery. A secondary concurrency test found that database-level unique constraints, not application logic, were responsible for preventing duplicate rows during simultaneous requests.

0 comments Read more at DEV Community

ProgrammingDEV Community ·

Graph Engineering Is the New Buzzword as AI Agentic Systems Grow More Complex

Tech circles are buzzing about 'graph engineering,' a framework for building AI systems where tasks are broken into specialised units called nodes, each handling a distinct role such as researching, writing, or fact-checking. These nodes are connected by edges that direct the workflow based on each node's output, allowing errors to be flagged and routed back for correction before moving forward. Developers familiar with tools like LangGraph note this approach to orchestrating AI agents is not new and has been in practice for over a year. The term itself only went viral recently, following a pattern of rebranding similar concepts — from context engineering to harness engineering to loop engineering. Analysts argue what is genuinely fading is the notion that a single model paired with a clever prompt can constitute a complete agentic AI system.

0 comments Read more at DEV Community

ProgrammingDEV Community ·

The Redistribution Gap: Why Existing Information Still Fails to Reach Everyone

Despite AI enabling faster content creation than ever before, a growing divide exists between information that is produced and information that actually reaches different audiences — a phenomenon termed the 'redistribution gap.' Factors such as social media platform algorithms, language barriers, content format, and paywalls all determine which version of information a person receives, if any. A researcher's detailed article, for instance, may never reach video-only consumers, non-English speakers, or free-tier users who see only partial content. AI tools now make it easier to redistribute the same core idea across multiple formats — articles, short posts, videos, audio, and translations — helping bridge these access gaps. The key opportunity in the AI era, the piece argues, lies not in creating more information but in strategically redistributing existing knowledge so it reaches the widest possible audience.

0 comments Read more at DEV Community

ProgrammingDEV Community ·

How One Missing Correlation ID Turned a Simple Bug Hunt into 90-Minute Ordeal

A software team received a customer-reported 500 error with only a timestamp to investigate, but their request had traveled through four separate services — an API gateway, an auth service, an orders service, and a payments service — each logging independently with no shared identifier. Tracking down the single failed request took roughly 90 minutes, requiring manual cross-referencing across four log stores and a database query to bridge a service that logged only an internal account key. The root cause of the debugging difficulty was the absence of a correlation ID: a single identifier generated at the entry point and propagated through every service hop and async message. The team retrofitted the system using the W3C trace context standard, ensuring the ID appeared in structured log fields across all services and fed into their tracing backend. Once full request visibility was established, the team discovered previously hidden issues — including an unknown retry loop doubling traffic and a cache that was missing far more often than assumed.

0 comments Read more at DEV Community