O 6.KR1
In order to create a latency SLO for realtime events, we need to first measure the latency in the realtime stream.
The difference between event_created and event_published gives the latency for the event.
To do
- Refine the realtime SDK, if needed, to capture the fields (event_created and event_published).
- Using the SDK, create a script that captures the events for 7 days, pushing timestamps (event_created and event_published) in a db. The script does not need to run continuously for a week. We can run it once a day for several hours keeping track of offsets of events. When we kill the script, we can capture the OS interrupt and dump the offsets in a file. In the next run, we can pass the latest offsets.
- Use the db data, calculate p50, p90, p99, p999 of latency
- Prepare a doc to share
Acceptance criteria
Latency SLOs are defined and shared.