Edge Data Replication unsteady

[thya-aerospace]

Description:

I tried to configure an edge device to sync a bucket towards a server bucket. For this purpose I followed the documentation here.

The main question that I would like to get answered is if this behaviour was already reported somewhere and I oversaw the solution? And to what extend the "data replication" feature really works or still is experimental?

Steps to reproduce:

1. Setting up remote connection on edge device
2. Setting up replication of bucket on edge device towards bucket on server
3. Fill bucket on edge device with local sensor data in a 1Hz frequency

Expected behavior: The bucket on the server shall show the same data as the bucket of the edge device over time.

Actual behavior:

• The data gets syncronized from the edge bucket to the server bucket only once when the server bucket is empty. Only a few data points. The data types are constanly the same.
• Then it can be seen that the "Current Queue Bytes" are growing until max configured queue value.
• The growing data in the bucket on the edge device then never gets synchronized with the server bucket again.
• The influxd logs of the server as well as of th edge device did not contain any message indicating a problem in that direction

Environment info:

• Influxd on Server
  • InfluxDB v2.6.1 (git: 9dcf880fe0) build_date: 2022-12-29T15:53:07Z
• Influxd on Edge
  • InfluxDB v2.6.1 (git: 9dcf880fe0) build_date: 2022-12-29T15:53:07Z
• Edge device has good and stable connectivity to the server

[jeffreyssmith2nd]

And to what extend the "data replication" feature really works or still is experimental? Edge Data Replication is a supported feature and is intended to work. However, it is fairly new and has not been battle tested in the way that the underlying storage layer has so it may not be as stable.

Just to make sure I understand correctly. When the server bucket is empty, the edge starts to replicate. At some point it then stops replicating and never restarts.

If you can query the replications through the API, that might show errors that logging does not.

Also do you have the ability to run a custom build on the edge? If so, I can give you a build with a firehose of logging turned on to see if that helps track down the issue.

[C-Aniruddh]

@jeffreyssmith2nd I am facing a similar issue, but in my case the remainingBytes are decreasing and I see a 204 success on replications.

However, when querying for data from the replicated server, there are less than 1% of points that are returned back.

I can run a custom build at the edge, if you can please forward the build to me!

[davidby-influx]

Replication error logging was improved significantly in 2.7.5 ad later versions.

See here for more information on how to debug replication problems: https://github.com/influxdata/influxdb/issues/24434#issuecomment-2219005265