I am seeing a mapping conflict on "message" field because in some older indices, "message" data type was set to "text" while in others it is "string." This is preventing me from using the field in any sort of analysis in Kibana due to the message "The type of this field changes across indices. It is unavailable for many analysis functions. The indices per type are as follows."
The problem is, all of the indices it shows as problematic on that page were deleted long ago. Obviously I can't reindex documents from an index that no longer exists. If those indices were deleted and they don't come up when I query the ESDB, how can I fix this?
Do you see any of these indices with GET_cat/indices?v
Can you verify the indices you're hitting the mapping exception on, the response should show you what index names are conflicting.
Do you have the example query you're using as well?
No, this is the problem--those indices were deleted in the past. None of the indices listed with conflicts actually exist. These are the indices listed as conflicting:
Do you have the mapping conflict error message you can show?
If you are getting errors back referring to indices that you cannot see are there, then I wonder if this might be a split brain scenario perhaps.
To detect this, check what each node in the cluster thinks it's node membership is with:
GET /_cat/nodes?v
And if you get different answers from any nodes then there might be a network partition.
Apache, Apache Lucene, Apache Hadoop, Hadoop, HDFS and the yellow elephant
logo are trademarks of the
Apache Software Foundation
in the United States and/or other countries.
