Hi everyone 👋

It's Friday again. 2020 has been a hell of a ride so far. I hope you're all well and surviving.

This week the focus of the team was on the continuation of the migration of our multi-arch builds from Jenkins to GitHub Actions, which means whole new workflows and the research & development of a build flow that can be reused for all network services (multi-arch and single arch). Work focused on implementing this for test.network builds. We’ve successfully built & deployed testnet Gateway from the new setup, and are now working on integrating the new Agent, Host and CDN builds into the existing deployment process.

We open sourced a new project that builds on earlier work from Adam to provide a bridge for GitHub Actions into Slack. You can see it here: https://github.com/edge/simple-slack-notify 🎉

Stargate session management was deployed to mainnet after a final successful (/stable) weekend on testnet.

Edge has now removed its reliance on Consul for session management; now tracks Host/Agent separately, which has allowed us to provide far better and fine grained data visualisation; and has vastly improved its multi-arch build ability.

It’s been a great couple of weeks.

The team also investigated a particular deployment issue with Gateway/Stargate. (Note that mainnet is stable and unaffected.)

We’re not yet 100% sure what the root cause is, but what we have been seeing on test.network is that on deployment of Gateway there are issues with it properly starting up on first boot, with the remedy seeming to be an auto restart of Stargate and then Gateway in sequence.

The working theory is that it is to do with the geo-lookup service, and some sort of race condition therein. We’re continuing to work on it, and won’t be releasing further updates to mainnet until it is resolved.

The team also solved a condition in mainnet whereby a DC partner had temporary limitations placed on some of its peering relationships, negatively impacting the performance of the network in some areas. The active routing of requests in the event of network slowdown at Gateway level prevents this sort of localised issue becoming a service interruption event.

And we've been running some Golang training sessions internally to help bolster our internal knowledge base.

A series of minor API and explorer fixes were also deployed. This involved changes to session management relating to how the sessions referred to the Stargate they were issued from.