Open YANG-DB opened 1 year ago
This can be useful, a few clarifying questions
The Y-Axis is the duration in milliseconds
Time-Slot structure Within each slot has a stacked list of spans collections that is organized according to its average duration. Each stacked span group height corresponds to the amount of spans in that group (relatively to the whole time-slot)
Open Traces dialog
The Y-Axis is the duration of parent span (sum duration of child pans)
Q:Assuming this is the sum of durations of spans, if there is one span that keeps timing out, would it affect the graph?
A: Parent spans which have timeout child spans are not included in the calculation
Q:What is a span collection?
Span collection is the list of span (parent spans) that are grouped to calculate the average latency Amount is the number of spans within the average range (similar to percentiles )
Q: Open Traces dialog
A: The traces table :
Is your feature request related to a problem? I would like to view the traces/spans in a timeline graph based projection to be able to observe trends and spikes
What solution would you like? As part of the diagnostic and investigation process I would like to view the traces / spans overlaid on a timeline to simplify the Observability and spikes appearance
Selection of graph span (categorical) dimension :
operation-name
/service-name
duration
in millisecondsbucket
of constant time slots, inside each slot are all the parent spans / traces that have completed within this time slot.Filter the traces using a dimension
Allow to filter the span graphs using a dimension filter
ERROR
filterTime-Slot structure
Each slot has a stacked list of spans collections that is organized according to its average duration and their belonging to any of the next percentiles : 50th (median), 75th, 90th, 95th, 99th, and 99.9th.
Each stacked span group height corresponds to the amount of spans in that group (relatively to the whole time-slot)
Fly-Over & Selection
A user can hover on each bar and see the fly-out of the next details
Average & percentile of that time slot
latency trend
Open Traces dialog :
navigate to the traces dialog using the selected time-slot, existing filters
Open discover dialog :
navigate to the logs discover dialog using the selected time-slot, existing filters and show all logs filtered
acts as traces to logs correlations functionality
What alternatives have you considered? None
Do you have any additional context?