Description
Currently, we start some redis spans as db.query spans and when we detect that they should be cache spans (driven by the used redis operation and user-provided cachePrefixes), we rename the span and adjust attributes.
This is a bit suboptimal for span streaming because users see their span as a cache.* span in the Sentry UI when in fact it was started as a db.query span by the SDK. This makes finding the right ignoreSpans filter very hard, given ignoreSpans is already applied at span start time.
Ideally, we could:
- start the cache spans right away as
cache.* spans and avoid as much post-start-processing as possible
- retain the same user logic (
cachePrefixes) to decide when a redis span should be started as cache
In a quick investigation, it seems like this is possible, but part of this issue is researching if my claim holds up :D
Description
Currently, we start some redis spans as
db.queryspans and when we detect that they should be cache spans (driven by the used redis operation and user-providedcachePrefixes), we rename the span and adjust attributes.This is a bit suboptimal for span streaming because users see their span as a
cache.*span in the Sentry UI when in fact it was started as adb.queryspan by the SDK. This makes finding the rightignoreSpansfilter very hard, givenignoreSpansis already applied at span start time.Ideally, we could:
cache.*spans and avoid as much post-start-processing as possiblecachePrefixes) to decide when a redis span should be started as cacheIn a quick investigation, it seems like this is possible, but part of this issue is researching if my claim holds up :D