Suranga Sandaruwan

When it stops, I'm getting started

[Suranga Sandaruwan @surangasw97]

@grok, would you enable Memory QoS throttling in production?

[Suranga Sandaruwan @surangasw97]

Would you stage npm releases for human approval?

[Suranga Sandaruwan @surangasw97]

Would you centralize resilience policies across AWS accounts?

[Suranga Sandaruwan @surangasw97]

@grok, would you restrict pull_request_target by default?

[Suranga Sandaruwan @surangasw97]

Would you dual-export metrics during an OpenTelemetry migration?

[Suranga Sandaruwan @surangasw97]

Will you pin ubuntu-24.04 before ubuntu-latest migrates?

[Suranga Sandaruwan @surangasw97]

@grok, which Kubernetes storage hardening control would you test first?

[Suranga Sandaruwan @surangasw97]

Would you enable AI Scan before CodeQL default setup?

[Suranga Sandaruwan @surangasw97]

Which Kubernetes metadata belongs on every telemetry signal?

[Suranga Sandaruwan @surangasw97]

@ardent__dev Google

[Suranga Sandaruwan @surangasw97]

@zaidmukaddam @claudeai @grok this is real?

[Suranga Sandaruwan @surangasw97]

@mayuri_3015 Exactly.

[Suranga Sandaruwan @surangasw97]

@arctic_hare_ @ChatGPT 🤣

[Suranga Sandaruwan @surangasw97]

GitHub disabled SHA-1 for HTTPS today across its cloud service and partner CDNs. Legacy browsers, API clients and Git-over-HTTPS tooling may now fail TLS negotiation; Enterprise Server is unaffected. Audit Git, OS TLS libraries and CI base images—not just developer laptops.

[Suranga Sandaruwan @surangasw97]

@fuyu_kasumii Yo

[Suranga Sandaruwan @surangasw97]

AWS ECS can now mark a rolling deployment successful before 100% of desired tasks are healthy. After that point, deployment rollback protection ends while remaining tasks scale normally.\n\nWhich production signal would you require before setting an early-success threshold below 100%?

[Suranga Sandaruwan @surangasw97]

AI coding agents need policy above the workspace.

GitHub lets enterprises centrally mark shell commands, file access and network domains as blocked, approval-required or allowed. Local settings cannot weaken those controls.

Which operation would you put behind approval first?

[Suranga Sandaruwan @surangasw97]

Google Cloud Monitoring now lets a chart override its dashboard’s time range.

That makes it easier to keep a 30-day capacity trend beside a 30-minute incident view—but it also puts two clocks on one screen.

What visual cue or review rule best prevents responders from treating those panels as time-aligned?

[Suranga Sandaruwan @surangasw97]

Agentic attacks are compressing the defender’s clock.

Google Threat Intelligence says one Q2 cloud compromise became an agent-enabled mass credential-harvesting campaign in under six hours.

Practical takeaway: shorten credential lifetimes, alert on unusual automation, and rehearse rapid revocation.

[Suranga Sandaruwan @surangasw97]

Kubernetes 1.37 moves gang scheduling and workload-aware preemption to beta. A distributed Job can preempt one member or evict the whole group together.

For production AI training, would you choose single or all—and which failure mode drives that choice?