-
QuestionI'm encountering two bugs with inconsistent behavior between Vector sinks that are preventing production deployment without forking:
I submitted PR #23427 to fix the null handling, but it was suggested we need a global fix across all sinks instead. The core issue: Sinks already behave inconsistently. Why block targeted fixes for user-impacting bugs while waiting for a comprehensive refactor that may not materialize soon? This forces production users to maintain forks just to work around known issues. Questions for the community:
Vector Configsinks:
loki:
type: loki
endpoint: "http://loki-gateway.monitoring.svc.cluster.local"
labels:
"*": '{{ loki_labels }}'
structured_metadata:
"*": '{{ structured_metadata }}'
request_id: '{{ request_id }}' # Becomes "<null>" if missing
encoding:
codec: json
except_fields: ["structured_metadata", "loki_labels"] # Ignored
elasticsearch:
type: elasticsearch
endpoints: ["http://victorialogs.monitoring.svc.cluster.local:9428/"]
encoding:
except_fields: ["structured_metadata", "loki_labels"] # Works correctly Vector LogsNo response |
Beta Was this translation helpful? Give feedback.
Replies: 1 comment
-
Hello @jmealo, I think my answer here #23040 (comment) addresses most of the questions raised in this discussion. Basically, the proper solution here is both the correct approach and also easier/cleaner to implement. Since we have many open discussions/issues on this already, I suggest continuing the discussion there to avoid fragmentation. |
Beta Was this translation helpful? Give feedback.
Hello @jmealo, I think my answer here #23040 (comment) addresses most of the questions raised in this discussion. Basically, the proper solution here is both the correct approach and also easier/cleaner to implement. Since we have many open discussions/issues on this already, I suggest continuing the discussion there to avoid fragmentation.