[APM] Service map: Show information for connecting lines between nodes (edge metrics)

[formgeist]

Summary

Related to #52869

Part of the information we want to display on the Service maps in order for users to investigate their performance issues further is the ability to see some basic metrics between two nodes in the map.

Solution

When the user hovers the connecting lines between nodes (service or connection), we will display a small tooltip element in the same place every time, placed on top of the other map content. This tooltip will display the following metrics;

• Avg. response time
• Requests per minute
• Error rate

It's worth noting that we'll not be supporting edge metrics for groups of connections like in this case;

It doesn't really make sense to aggregate at such a level and adding individual metrics in the popover list will also be a challenge.

[elasticmachine]

Pinging @elastic/apm-ui (Team:apm)

[formgeist]

@roncohen @nehaduggal I'm wondering where the implementation challenge lies within APM that would enable us to serve the inter-connection metrics that we talked about last sync meeting? I know we're not focusing on it right now but, corrected me if I'm wrong, I believe this is something that is key to our service map feature in the short-term.

[smith]

Most of the work for the front-end is done in #52870, but we're pushing this back to 7.8 so we can have the data we need.

[felixbarny]

Looks great!

Some feedback:

• Having error rates would be a great help to identify failing connections.
• Although having the edge stats on the same space on the screen may make it more consistent, it requires jumping back-and-forth quite a bit: Focus is first on mouse pointer and trying to hover over a connection, then focus is at the bottom of the screen. When wanting to inspect an adjacent connection, focus jumps back up and down after the pointer is on the other connection.
• A visual indicator like line thickness could help to get a glance of whether there's low, medium or high traffic between services.

[formgeist]

Thanks for the feedback @felixbarny - I have some comments and possible background for why we chose to go with a static place for the information.

Having error rates would be a great help to identify failing connections.

I don't think this is on our radar, but it should be. I know there's an issue open for adding destination service metrics to the agents, so I'll add a comment there.

Although having the edge stats on the same space on the screen may make it more consistent, it requires jumping back-and-forth quite a bit: Focus is first on mouse pointer and trying to hover over a connection, then focus is at the bottom of the screen. When wanting to inspect an adjacent connection, focus jumps back up and down after the pointer is on the other connection.

While one of the reasons was to keep a consistent place for the tooltip information, we already experimented with adding the information on the line but because of the changing space around nodes and connecting lines, and that they're also drawn in various directions, made for a hard UX as well. Seeing as we have changed our layout a bit since we initial designed this, I'll go back to the drawing board to explore whether there might be a better way to have the information placed by the connecting line, closer to the context. I'll open issue to track.

A visual indicator like line thickness could help to get a glance of whether there's low, medium or high traffic between services.

Yeah, we discussed having this, but we need an issue to track, so I'll create for this as well.

Thanks again 👍

[formgeist]

I've updated the description with the new design solution for displaying these edge metrics per connection. We have another use-case where connections are grouped when there's more than 5 for a given node. We'll for the time being not support aggregated metrics for those connections, nor show it in the popover.

@nehaduggal Do you think we should prioritize a solution for displaying metrics even for grouped connections? One option would be to sort the list of connections by either latency, traffic, or error rate and display the metrics per connection in another way e.g. a flyout or similar.

[botelastic]

This issue has been automatically marked as stale because it has not had recent activity. It will be closed if no further activity occurs. Thank you for your contributions.

[alpdud]

hi All, Tatacliq is one of the 3rd largest fashion/beauty etailer in IN. They're currently using Dynatrace for a limited no of apps & logging is via ELK OS. Due to cost, billing model and other limitations they're keen to explore other options. At present their SREs are tuned to monitor app traffic/high level flow through service map. While custom dashboard options suggested but their team is insisting on having traffic flow (Transcations) superimposed on APM service map. This is exactly the enhancement they're looking for - hence checking if it's planned to be prioritised any soon.

[Erikg346]

+1
Just like @alpdud, this has been asked as other APM tools tend to fulfill this requirement.