Repository navigation
Replies: 8 comments 7 replies
@VA7ODR, Welcome to Meshtastic! 👋Thanks for opening your first issue. If it's helpful, an easy way If you have ideas for features, note that we often debate big ideas We're very active on discord, Welcome to the community! ❤️ |
|
A couple points based on my opinion:
|
|
My 6 solar remote routers are suffering under this restriction right now:
|
|
Couldn't the node store data—for instance, hourly readings from the last 24 hours—and make them available upon the owner's request? Meanwhile, the standard 12-hour interval could remain in place. |
|
If anything routers should be the one role that is an exception for telemetry. Routers are critical nodes on any mesh and as such monitoring their health is important. Sure, administrators can send telemetry requests multiple times a day but we all know that will just lead to scripted floods of polling requests. Not ideal vs the old system of simply passively collecting data. On our mesh, which spans more than 500km, some mountain nodes are only accessible for a month out of the year. Monitoring their telemetry for power and environmental conditions is vital to understand what is going on up there and what is needed on the only annual maintenance trip that is feasible. Once or twice a day is insufficient to identify odd solar charging behavior, anomalous temperature spikes, etc. On one of our routers on a remote island mountain top we are seeing a very odd pattern of the node periodically not charging. Thanks to our long term data collection via Home Assistant we were able to remotely determine the likely cause is thermal expansion on some component of the charging system. This kind of deduction is impossible without reliable telemetry multiple times a day. 4 hours is a good compromise but I argue it should be 12 hours default and owners can set to whatever they want, as before. Routers are a teeny tiny percentage of the nodes on any given large mesh and their telemetry should be prioritized over all other types. Including specialty roles like tracker, sensor. Without healthy routers the mesh does not mesh long distance. |
|
One of our hosting arrangements is to enable high school kids to get the sensor data from the node (temp, pressure, etc). Just adding to the examples above showing that more frequent data is sometimes useful. |
|
Extending the telemetry transmission intervals has led to the following issues:
It would be better to maintain the 4-hour telemetry interval while limiting the number of hops—for example, to a maximum of 2—so that the telemetry reaches a node capable of forwarding data to MQTT. Naturally, if network traffic increases (e.g., by 40%), this hop limit could be reduced by 1. From what I have observed, most users operating nodes with the ROUTER or ROUTERLATE role use their own custom builds. Introducing restrictions only creates chaos when it comes to monitoring remote nodes. Furthermore, most users with these roles specifically include a custom feature to broadcast information about their neighbors over LoRa. It would be beneficial to implement a standard mechanism for broadcasting neighbor information—including SNR values—limited to, say, a maximum of 1 hop. |
|
Currently fixed routers send their position 4x a day and telemetry twice a day. Surely it is more important to send telemetry than a fixed position so i propose at the very least swapping telemetry to 4x a day and position to 2x. |
Uh oh!
There was an error while loading. Please reload this page.
Category
Other
Hardware
Rak4631
Is this bug report about any UI (https://meshtastic.org/docs/configuration/device-uis/) component firmware?
Firmware Version
2.7.26
Description
Router telemetry 12-hour hard floor on default channels breaks remote infrastructure monitoring
Summary
Since #9937,
ROUTERandROUTER_LATEnodes participating in a default channel have their configured telemetry intervals coerced to a minimum of 12 hours:NodeDB::NodeDB()applies this wheneverchannels.hasDefaultChannel()is true usingDefault::getConfiguredOrMinimumValue().I understand the motivation behind #9937: infrastructure roles were not participating in normal congestion-based telemetry scaling, so a coarse minimum was introduced to prevent routers from contributing excessive routine traffic to very large meshes.
The problem is that this turns a congestion-management policy into a global role-based policy. A router on a quiet rural mesh is subject to exactly the same 12-hour minimum as one in a thousand-node metropolitan mesh, regardless of actual channel utilization, airtime utilization, modem preset, mesh size, configured interval, or topology.
I would like to propose replacing the unconditional 12-hour floor with congestion-aware behaviour, or providing an explicit strict/high-congestion mode that enables the current behaviour.
Why the current behaviour is problematic
1. Router telemetry is operational infrastructure telemetry
Routers are disproportionately likely to be unattended, solar-powered, difficult to access, and installed at remote elevated sites.
Device and power telemetry are therefore not merely informational. They are how operators detect:
Two telemetry samples per day are often insufficient to diagnose these conditions.
For example, with a solar node, a 12-hour interval can easily miss most of the useful portion of the charge/discharge cycle. A node can deteriorate substantially between samples, and once the node stops transmitting there is no way to retrieve the missing history.
This exact use case has previously been raised in #4466, where infrequent telemetry from remote mountaintop routers made it difficult to detect under-charging and moisture ingress.
2. It conflicts with Meshtastic's concept of an "online" node
Meshtastic defines an online node as one heard within approximately the previous two hours.
At the same time, infrastructure nodes on a default channel may now be prevented from originating telemetry more frequently than every twelve hours.
Telemetry is not the only packet that can update node freshness, but these policies nevertheless work against one another. A quiet, healthy infrastructure node can go through substantial periods where no routine router-originated state packet is available to demonstrate that it is still healthy.
This makes "online only" node filtering substantially less useful for infrastructure monitoring.
An infrastructure operator looking at the node list should be able to distinguish:
Making infrastructure nodes deliberately quiet for periods much longer than the application's freshness window makes those states unnecessarily ambiguous.
3. Explicit configuration is silently overridden
The applications expose telemetry reporting intervals as configurable values.
An operator may deliberately configure device or power metrics for 30 minutes, 1 hour, or 2 hours because that cadence is appropriate for their deployment.
For a router participating in a default channel, firmware can replace that choice with 12 hours.
That is surprising behaviour because the configured value no longer describes the effective behaviour of the node.
If firmware must override a configured interval for network safety, the effective interval should ideally be visible to the user and the reason for the override should be observable.
4. Default-channel membership is not a congestion measurement
The current test is effectively:
But participating in the default/public mesh does not imply that the mesh is congested.
There are many rural and regional deployments where:
Those meshes pay the same telemetry penalty as extremely dense urban networks despite not having the condition the policy is intended to mitigate.
5. Avoiding the default channel is not a good workaround
A network operator can theoretically avoid this behaviour by not participating in a default channel, but doing so sacrifices interoperability with the surrounding public Meshtastic network.
Congestion control should not require choosing between useful infrastructure telemetry and participation in the public mesh.
Suggested approach
I am not proposing simply removing congestion protection.
The congestion problem #9937 was intended to address is real. I am suggesting that the protection be made conditional on congestion rather than on role and default-channel membership alone.
Meshtastic already has several useful inputs for making this decision:
One possible approach would be:
For example, a quiet mesh could retain the existing normal telemetry minimum, while progressively increasing the effective interval as mesh size/channel utilization rises, eventually reaching the current 12-hour infrastructure limit in genuinely congested conditions.
Another possibility would be an explicit congestion-management policy such as:
where:
The exact API is less important than avoiding an unconditional global assumption that every public-channel router operates in a highly congested mesh.
Another possible mitigation: reduce reach rather than frequency
If frequent infrastructure telemetry is undesirable across a large mesh, reducing its propagation distance may be preferable to suppressing the telemetry at its source.
For example, a router could originate device/power telemetry at the configured interval but have routine telemetry automatically use a smaller hop limit under congestion.
That would allow nearby infrastructure-monitoring nodes to retain useful health data without sending every telemetry sample across the entire extended mesh.
This is similar in spirit to the ideas already being discussed in #6675.
Suggested minimum acceptance criteria
I would consider the issue resolved by any design that provides these properties:
Relevant history
The goal here is not to increase unnecessary traffic on already congested meshes.
It is to make congestion control proportional to congestion, so that infrastructure on quiet and rural meshes does not lose valuable health-monitoring functionality because of conditions that do not exist on that mesh.
Relevant log output
All reactions