input.modbus TCP connection handling

[mihailescu2m]

It seems that after successfully gathering data over modbus TCP, the connection is not released.
This makes the telegraf plugin have unique access to the modbus device, and other connections fail even if telegraf is not actively gathering data but it's idle waiting for the next interval.
Can this be fixed?

[ssoroka]

definitely! which modbus plugin are you using?

[mihailescu2m]

I am using the default telegraf plugin

example config:

[[inputs.modbus]]
  name = "inverter"
  slave_id = 1
  interval = "10s"
  timeout = "4s"
  controller = "tcp://inverter:1502"
  holding_registers = [
    { name = "dcPower",           byte_order = "AB",   data_type = "INT16",  scale=1.0, address = [40100]},
    ...

[ssoroka]

It looks like it defaults to disconnecting after a minute of no activity, which seems like a reasonable default. Are you seeing any problems with connections left open longer than that, or issues with the connections not being used?

[sjwang90]

@mihailescu2m ping

It looks like it defaults to disconnecting after a minute of no activity, which seems like a reasonable default. Are you seeing any problems with connections left open longer than that, or issues with the connections not being used?

@mihailescu2m

[mihailescu2m]

The problem is when I am trying to access the same device over modbus from another software.
So let's say every minute at :00 and :30, telegraf gathers data (~3 seconds duration), and I want another software, also every minute at :15 and :45 to access modbus (for ~3 seconds). This is impossible, since basically telegraf never releases the connection (and there can be only 1 client connection)

[rwalli]

I can confirm the bug with tcp-timeout and in my case i figured out that telegraf close the TCP-connection after the “intervall-setting”.

[srebhan]

@mihailescu2m telegraf definitively does not close the connection after the query. A pull-request adding a disconnection_after_query (or some less-bulky name) taking a boolen or, even better reconnect_after_queries taking a number of queries (zero = do not disconnect) is welcome.

[srebhan]

@mihailescu2m and @rwalli I implemented a disconnection setting in PR #9256. It would be nice if you can test the new setting. You can use the artifacts built by CI and add the following to your modbus-plugin config:

workarounds = {
  close_connection_after_gather = false
}

Please let me know if you find any issue. Thanks you in advance!