We experienced occasional duplicate log entries in application insights
that we were able to trace back to failed transmissions:
An error occurred while sending the request. | Unable to read data from the transport connection: Connection reset by peer. | Connection reset by peer

The ServerTelemetryChannel, which we are using, will then store the transmission in the file system and retry after a certain interval. The issue is that the first transmission was also successful and a networking issue happened for some unknown reason afterwards. So the data gets processed twice.

To prevent this, I would propose the following solution:

• Add an id/digest property to the API that is receiving log transmissions


• implement a logic, that will reject a transmission, if a transmission with the same id has already been processed earlier for the same instrumentation key.

This will allow the ServerTelemetryChannel SDK to implement a logic where the id remains constant for one transmission and its retries, retaining the ability to recover a transmission but without generating duplicate logs

Current Behavior: API gateway is shown with its mapped backends and requests from clients(Web) to API Management Gateway through WAF is shown as HTTP calls and are shown separately.

Expected Behavior: When we see a HTTP Role Type in Application map like say api.contoso.com, The application map search and linking process can consider suggesting expansion of that role to the mappings of WAF in an application map. It is mostly possible this is already mapped to a backend pool in WAF and these can be shown easily. This will make sure the application map created at resource group level is also automatically picking up useful clues about the linking of the apps behind an API Management gateway.

In Failures blade, observed an endpoint with approx 2000 failures in last 24hrs. When I click on the operation name in the table, on the right side a list of operations/calls to that end point are listed. When I click on an operation, the end-to-end transaction panel is shown, but instead of showing just the selected operation it starts loading every call to that endpoint. It doesn't even limit it to the last 24 hours. Instead of being able to see a single opeartion it starts loading (and keeps loading), so far it is over 20 thousand! and it is impossible to use the page due to performance degradation. Making this feature useless. Please provide a fix or an alternative.

After making a search and then selecting a time frame in the UI there's no way, that I could find, to reset the search to the current time frame.

Repro:
Do a search over the last 30 minutes
Using the UI select a portion of that 30 minutes
After reading the results try to get the latest 30 minutes again

Results:
UI doesn't have any way of clearing the time
Even the 30 minutes I was last at is not selectable in the UI
In order to get the 30 minutes back I need to select the last hour, hit apply and then select last 30 minutes and apply

Expected Results:
I should be able to clear the timeframe in the UI to go back to the last time range I was at.

Support the ability to qualify / bucket a dependency / span trace based on the the names of the inner dependencies / spans.

We have a program that sends dependencies to AI via OpenCensus spans. The program has a long path and a short path. The longer path is an order of magnitude loger in time that the snorter path.

[1[2,3,4]] and [1[2,3,4,5,6]]

We want to profile both the short path and the long path. Today they are are all statistically mixed because the outer span name is the same. We can't set the outer dependency/span name prior to entering the code because of the way the code decides whether to run long or short path.