Rebuilding Elasticsearch Index

On this page

  1. Problem
  2. Solution
    1. Scale down the Elasticsearch and CluedIn
    2. Delete the Elasticsearch persistent volume
    3. Scale up the Elasticsearch
    4. Scale up CluedIn
    5. Reprocess the entities

Problem

Elasticsearch index is corrupt or lost. How can I save my data?

Solution

Don’t worry; your data is not lost and still exists in Neo4j and Microsoft SQL Server databases. You can restore the Elasticsearch index by reprocessing CluedIn entities.

If the index is completely broken, it can be easier just to clear the disk and build a new index.

In this article, we will show you how to do that. Please, keep in mind that the operations listed below are dangerous. Therefore you should be careful and understand what these operations do.

The commands also depend on your setup and can be slightly different. Further, we assume that CluedIn is installed with a standard Helm chart in a cluedin namespace.

Scale down the Elasticsearch and CluedIn 

kubectl scale deployments --selector app=cluedin --replicas=0 -n cluedin;
kubectl scale statefulsets --selector app=elasticsearch-master --replicas=0 -n cluedin;

After this step, you should not see any Elasticsearch or CluedIn -server, -processing, or -crawler pods.

Delete the Elasticsearch persistent volume 

kubectl get pv

You will see a persistent volume with a claim named like default/elasticsearch-master. Remember its name.

Delete the Elasticsearch persistent volume:

kubectl delete pv default/elasticsearch-master -n cluedin

Scale up the Elasticsearch

It’s important to do it before you scale up CluedIn deployments.

kubectl scale statefulsets --selector app=elasticsearch-master --replicas=1 -n cluedin

After this step, a new persistent volume will be created.

Scale up CluedIn 

kubectl scale deployments --selector app=cluedin --replicas=1 -n cluedin

You will see that CluedIn is running, but there are no entities in the UI.

Reprocess the entities

Now, you can reprocess the entities. However, to not overload the queue, we recommend doing it separately for each entity type:

curl \
  --location \
  --request GET '{url}/api/api/admin/commands/process/all/entityType/?organizationId={orgid}&entityType={entitytype}&minsize=0' \
  --header 'Content-Type: application/json' \
  --header 'Authorization: Bearer ...

For more details about the API calls, please refer to the API documentation.