When OPENSEARCH_BATCH_WRITE_REFRESH_DEFAULT = "true", the client will trigger a global refresh command. this forces a refresh of all indexes, even read-only ones, which results in an error. A refresh should be per index. This default setting is dangerous and was initially added to make the client more debuggable. It would be a better experience if the default were set to false, so as to remove this latent bug that eventually causes issues.
Hi, This issue happened to us when using default settings of hadoop-client which sets the refresh as "true". Pls set the default to false so write operations into index do not fail.
What is the bug?
When OPENSEARCH_BATCH_WRITE_REFRESH_DEFAULT = "true", the client will trigger a global refresh command. this forces a refresh of all indexes, even read-only ones, which results in an error. A refresh should be per index. This default setting is dangerous and was initially added to make the client more debuggable. It would be a better experience if the default were set to false, so as to remove this latent bug that eventually causes issues.
https://github.com/opensearch-project/opensearch-hadoop/blob/203d53f3f5c052c27691a079f684433f51035f02/mr/src/main/java/org/opensearch/hadoop/cfg/ConfigurationOptions.java#L97