[ML] Creating a new job fails if it's using a reindexed results index #38773
Description
Steps to reproduce:
- Install ES 5.x
- Run a job
- Do a full cluster restart update to ES 6.7
- Install Kibana 6.7
- In the migration assistant reindex all the indices created in 5.x into the 6.x format
- Try to create a new job
The following error is the result:
{
"error" : {
"root_cause" : [
{
"type" : "invalid_index_name_exception",
"reason" : "Invalid index name [.ml-anomalies-shared], already exists as alias",
"index_uuid" : "_na_",
"index" : ".ml-anomalies-shared"
}
],
"type" : "invalid_index_name_exception",
"reason" : "Invalid index name [.ml-anomalies-shared], already exists as alias",
"index_uuid" : "_na_",
"index" : ".ml-anomalies-shared"
},
"status" : 400
}
After reindexing the output of _cat/aliases is this:
alias index filter routing.index routing.search
.kibana .kibana_1 - - -
.ml-state .reindexed-v6-ml-state - - -
.ml-state-write .reindexed-v6-ml-state - - -
.security .security-6 - - -
.ml-anomalies-.write-farequote .reindexed-v6-ml-anomalies-shared - - -
.ml-anomalies-farequote .reindexed-v6-ml-anomalies-shared * - -
.ml-anomalies-shared .reindexed-v6-ml-anomalies-shared - - -
.ml-notifications .reindexed-v6-ml-notifications - - -
.ml-annotations-read .ml-annotations-6 - - -
.ml-annotations-write .ml-annotations-6 - - -
Instead of checking whether the required indices exist, our code needs to check whether the required name exists as an index or alias. It should tolerate opening a job when .ml-anomalies-shared is an alias rather than an index.
It is critical that we fix this before 6.7.0 GA.
Metadata
Metadata
Assignees
Labels
Type
Fields
Give feedbackNo fields configured for issues without a type.