Skip to content
GitHub

HotLoop Flow 2.16.0 release notes

HotLoop Flow 2.16.0 lets one flow file, and so one set of FlowLab diagrams, be pushed to every Flow install a tenant runs in EmberNet, with a check after each one, and a failed one put back. It also lets two installs share one job, with a lease that makes sure only one of them writes to equipment at a time. It is amd64 first, with arm64 following in 2.16.1.

  • Push to a tenant’s stations (#201). hotloop-flow stations list -tenant ID and hotloop-flow stations push -tenant ID -file flows.json put one flow file on each Flow install EmberNet shows as running for the tenant, one at a time. The stations come from the EmberNet dashboard and are never typed in, and a row of any other tenant is dropped. After each deploy the station must answer the revision every station is to have, have no node that failed to start, and have every diagram of the file that starts by itself present and running or finished. A station that fails is rolled back to the deployment it was on and the push stops; the report says which stations were done, failed, skipped and never started. A push that would restart a running logger changes nothing until the operator says yes at the prompt, passes -confirm-restart, or gives a -window that holds now. docs/stations.md is the page.
  • The same push over the API and MCP (#201). GET /stations and POST /stations/push, and the MCP tools list_stations and push_to_stations, which brings the tools to 43. They are there only where the four stations settings are set.
  • A warm standby with a fencing lease (#204). Two Flow installs can share one job. One PostgreSQL row holds a lease and a number that rises each time the holder changes. The holder checks the lease right before every write to equipment, in the Modbus, S7, OPC UA and EtherNet/IP write nodes, the HotLoop script and recipe nodes, and the FlowLab diagram writes to HotLoop, instruments and serial ports. A standby, or a holder that has lost its lease, has its write refused with a counted error before anything is sent, and the standby takes the lease over after the active one’s lease runs out. GET /failover and the MCP tool get_failover, which brings the tools to 44, say whether this install is active, standby or lost, who holds the lease and how many writes were held back. The editor shows a banner on a standby, in the 13 languages, and docs/failover.md is the page.
  • Chart values and Quadlet lines for it (#204). failover.urlSecret, failover.name, failover.ttl, failover.renewEvery and failover.writeMargin in the chart, with the Secret holding the PostgreSQL URL mounted as a file, and the environment variables and Podman secret for a Quadlet unit in docs/deploying.md.
  • Chart values and Quadlet lines (#202). stations.dashboardUrl, stations.tenant, stations.dashboardTokenSecret and stations.deployTokenSecret in the chart, with the two Secrets mounted as files, and the four environment variables and two mounted secrets for a Quadlet unit in docs/deploying.md.
  • A start whose failover settings could let the active run short of lease between renewals, such as a write margin as long as the lease, is refused.
  • A start with only some of the four stations settings is refused, and so is a chart render with some of them. All four or none.
  • A token that pushes to stations needs status.read as well as flows.read and flows.write, because the check after each deploy reads the station’s diagrams. hotloop-flow token now prints that, and so do the README and the stations page. A plain deploy token still works for deploy as before.
  • The revision a flow set will have once saved can now be known before it is saved, from one function the save also uses. Nothing about revisions changed.
  • The docs name hotloop-flow-mcp-server where they said mcp-server (#197, #199, #200).
  • The static binary is 51,859,618 bytes (49 MiB), under the 50 MiB budget in ci.yml.

stations.dashboardUrl, stations.tenant, stations.dashboardTokenFile and stations.deployTokenFile in the config file, or HOTLOOP_FLOW_STATIONS_DASHBOARD_URL, HOTLOOP_FLOW_STATIONS_TENANT, HOTLOOP_FLOW_STATIONS_DASHBOARD_TOKEN_FILE and HOTLOOP_FLOW_STATIONS_DEPLOY_TOKEN_FILE. All empty by default, which leaves the feature off. In the chart they are the stations.* values above.

failover.urlFile, failover.name, failover.holder, failover.ttl, failover.renewEvery and failover.writeMargin in the config file, or HOTLOOP_FLOW_FAILOVER_URL_FILE, HOTLOOP_FLOW_FAILOVER_NAME, HOTLOOP_FLOW_FAILOVER_HOLDER, HOTLOOP_FLOW_FAILOVER_TTL, HOTLOOP_FLOW_FAILOVER_RENEW_EVERY and HOTLOOP_FLOW_FAILOVER_WRITE_MARGIN. The name defaults to flow, the holder to the host name, the ttl to 15 seconds, the renewal to a third of it and the write margin to 5 seconds. With no URL file the feature is off and every write goes out as before. In the chart they are the failover.* values above.

A plain helm upgrade to the 2.16.0 image and chart, or the new tag in your Quadlet unit. Nothing happens on first start and nothing needs migrating. A flow file loads and saves byte for byte as before. This release is amd64 only: arm64 boxes stay on 2.15.1 until 2.16.1, which is the same code with the arm64 image added.

Image: ghcr.io/hotloop-io/hotloop-flow:2.16.0, linux/amd64 first, with arm64 following in 2.16.1. Chart: https://hotloop.io/hotloop-flow/. The README at this tag has the rest, with pictures: https://github.com/HotLoop-io/hotloop-flow/blob/v2.16.0/README.md