Closed jpkrohling closed 9 months ago
@jpkrohling I would like to work on this issue. I am new to go, would you please guide me to which files should I look into. It would be really helpful 😅.
I am not clear what this issue is about, it needs more details. It sounds closely related to #166, which is unlikely a "good first issue".
@knrt10 I would recommend picking another ticket among those labeled good first issue.
Ok thanks let me try.
@yurishkuro this issue is only about adding a sort clause to the query used by the UI endpoint to list the traces. The in-memory might not be affected, but it would be good to make sure both the Cassandra and ES plugins have such clause in place.
As I understand, #166 is about traces that won't "ever" appear in the search results. This here is about the search results being different at each refresh. Trace "123" might show up the first time the search is executed, but then not show up when the page is reloaded, showing up again on a third reload.
the fundamental issue with Cassandra search may cause different order as well. The UI already has a Sort control, so if that's not helping then the issue is with the backend storage returning non-stable results, and I think it's mostly the problem in #166.
We've seen this with the all-in-one version of Jaeger (agents deployed as a DaemonSet) when the number of total traces is larger than the number of desired traces to return.
I've seen this with ES as the backend storage on Jaeger 1.40.0. Is there any update for this issue?
We do not make stable results order a requirement for backend storages, not should we imo, given that this is observability data, not systems of record, and the trade-offs are made in favor of recency and availability, not consistency.
Requirement - what kind of business use case are you trying to solve?
Problem - what in Jaeger blocks you from solving the requirement?