When processing http outgoing metrics, current sink will search for the host name of target ip. If there're lots of http outgoing requests with different ip addresses, there will be lots of queries on these ip.
This causes too many spans in the log for one transaction.
Cause
If a item is missed, it will be added into a cache with life time up to 1 minutes.
So for the first large batch, all items miss, and leads to above problem.
Solution
The CachableMetaStorage needs to be updated to add a periodic routine to load app names from underlying metadata storage into a cache. And the query routine always load from cache.
Description
When processing http outgoing metrics, current sink will search for the host name of target ip. If there're lots of http outgoing requests with different ip addresses, there will be lots of queries on these ip.
This causes too many spans in the log for one transaction.
Cause
If a item is missed, it will be added into a cache with life time up to 1 minutes. So for the first large batch, all items miss, and leads to above problem.
Solution
The CachableMetaStorage needs to be updated to add a periodic routine to load app names from underlying metadata storage into a cache. And the query routine always load from cache.