Open jeqo opened 5 months ago
Currently, fetch.max.wait.ms is reused on remote fetches to cancel operations if taking longer than timeout. Given the default value is 500ms, it may end up being too small for the whole remote operation to complete.
fetch.max.wait.ms
Ideally this would be default by a different configuration. Some alternatives are being discussed here: https://issues.apache.org/jira/browse/KAFKA-15776
Once a better approach for this issue is implemented upstream, update the plugin to support it properly.
https://cwiki.apache.org/confluence/display/KAFKA/KIP-1018%3A+Introduce+max+remote+fetch+timeout+config+for+DelayedRemoteFetch+requests is accepted and targeting 3.8 -- once that's released, we can close this issue.
Currently,
fetch.max.wait.ms
is reused on remote fetches to cancel operations if taking longer than timeout. Given the default value is 500ms, it may end up being too small for the whole remote operation to complete.Ideally this would be default by a different configuration. Some alternatives are being discussed here: https://issues.apache.org/jira/browse/KAFKA-15776
Once a better approach for this issue is implemented upstream, update the plugin to support it properly.