←
Ravion1,002 posts
Ravion avatar

Ravion

@ravion

Build a faster, safer software factory. Powerful building blocks for CI/CD pipelines, deploys, and Terraform automation, in your own cloud.

◷ Joined July 2021
Ravion avatar
Ravion@ravion·Aug 6

Hello world. All systems are go!

🔥 ANNOUNCING RAVION 🔥 The command center for your cloud infrastructure For startups: the easiest, most powerful way to self-host in your cloud For scaleups: a framework for internal developer platforms Result of 14 months. 1M LoC. Everything we learned from Flightcontrol.

1:55
Ravion avatar
Ravion@ravion·4h

Shared NAT gateways finally get an itemized bill. New Networking Flexible Cost Allocation splits NAT gateway and VPC peering data charges by the subnet, VPC, account, or instance that sent the traffic, right in your CUR. No Flow Logs scripts. Free. aws.amazon.com/about-aws/wha…

Post attachment
Ravion avatar
Ravion@ravion·21h

Ravion pipelines now cost significantly less to run, including GitHub Actions jobs on Ravion runners. Jobs are now scheduled and placed on hosts much more efficiently, and we fixed bugs in that placement system. ravion.com/docs/github-actio…

Post attachment
Ravion avatar
Ravion@ravion·Nov 17

Graviton5 just reached managed Postgres and MySQL. R9g is GA on Aurora and RDS. On Aurora, AWS cites up to 30% better performance and up to 22% better on-demand price-performance vs Graviton4. Four Regions for now. Check engine versions. aws.amazon.com/about-aws/wha…

Post attachment
Ravion avatar
Ravion@ravion·Nov 16

Multi-account AWS teams can stop sharing KMS keys just to share a secret. Secrets Manager now replicates across accounts. One team owns and rotates the secret; another account gets a synced local copy with its own policy and KMS key. Same Region only. aws.amazon.com/about-aws/wha…

Post attachment
Ravion avatar
Ravion@ravion·Nov 16

CloudWatch Logs retention is finally an account-level setting. One policy covers every log group, or every group under a prefix like /aws/lambda, including new ones. Careful: it overrides longer per-group retention. Dry run first. aws.amazon.com/about-aws/wha…

Post attachment
Ravion avatar
Ravion@ravion·Nov 13

Read your module logs from the terminal with Ravion CLI 1.33.0: ravion module log-sources <module-id> ravion module logs <module-id> --tail --jsonl CloudWatch or EKS Loki. For scripts: finite --json returns one array of batches; streaming uses --jsonl. ravion.com/docs/modules/logs

Post attachment
Ravion avatar
Ravion@ravion·Nov 13

AWS now has its own local emulator, free to use. Your existing CLI and SDK calls hit localhost and return in milliseconds. No deploy, no AWS account. Great for coding agents. It doesn't evaluate IAM policies, so test security on real AWS. aws.amazon.com/about-aws/wha…

Post attachment
Ravion avatar
Ravion@ravion·Nov 11

Custom pipeline steps in Ravion now support watch_paths. If no changed file matches since the step's last successful run in the same variant, the step is skipped and the steps after it keep going. API tests don't rerun when only the docs changed. ravion.com/docs/pipelines/co…

Post attachment
Ravion avatar
Ravion@ravion·Nov 11

Ravion pipeline steps can now declare `needs`. List the earlier steps a step depends on and it starts as soon as those finish, alongside everything else. needs: [] starts it right away. If a step it needs fails or is skipped, it's skipped too. ravion.com/docs/pipelines/co…

Post attachment
Ravion avatar
Ravion@ravion·Nov 10

Ravion pipelines now support matrices. Define matrix.axes on a step and it runs once per combination: 2 Postgres versions × 3 Node versions = 6 copies. Tune it with include, exclude, max_parallel, and fail_strategy. ravion.com/docs/pipelines/co…

Post attachment
Ravion avatar
Ravion@ravion·Nov 10

Ravion's EC2 Service can now move running instances to a new AMI in place. Instance IDs, network interfaces, IPs, and data volumes stay, and attached volumes are snapshotted first. The root disk is rebuilt, so hosts without a data volume need consent. ravion.com/docs/module-defin…

Post attachment
Ravion avatar
Ravion@ravion·Nov 9

Stop guessing CPU and memory for pipeline steps. Set cpu, memory, or disk to auto on an interruptible sandbox step and Ravion sizes it from the step's earlier runs. Out of memory or disk? It gets more and retries in the same run. In preview. ravion.com/docs/concepts/san…

Post attachment
Ravion avatar
Ravion@ravion·Nov 9

Small inference jobs no longer need a whole GPU each on ECS. Multiple tasks can share one NVIDIA GPU through MPS, with per-container memory limits. EC2 + Managed Instances, not Fargate. Software sharing, not hardware isolation. aws.amazon.com/about-aws/wha…

Post attachment
Ravion avatar
Ravion@ravion·Nov 6

ECS Managed Instances can now pack tasks from several underused instances onto fewer machines, including Spot. Previously, it evaluated each instance alone. Now it can optimize the fleet together. Automatic, with no extra fee. aws.amazon.com/about-aws/wha…

Post attachment
Ravion avatar
Ravion@ravion·Nov 6

Hitting Aurora PostgreSQL serverless's 256-ACU ceiling? The limit is now 384 ACUs: 50% more headroom, roughly 768 GiB of memory per instance. Requires Platform Version 4 and a supported engine version. MySQL stays at 256 ACUs. aws.amazon.com/about-aws/wha…

Post attachment