We have a working OpenSearch cluster in dse-k8s-eqiad:

kubectl get po
NAME                                                              READY   STATUS      RESTARTS   AGE
cluster-masters-0                                                 1/1     Running     0          157m
cluster-masters-1                                                 1/1     Running     0          152m
cluster-masters-2                                                 1/1     Running     0          151m
cluster-securityconfig-update-w64t5                               0/1     Completed   0          157m
operator-opensearch-operator-controller-manager-54cc5d7965flgrc   1/1     Running     0          2d2h

The next steps will be to work on TLS and ingress.

Change #1196412 had a related patch set uploaded (by Btullis; author: Btullis):

[operations/deployment-charts@master] Deploy the opensearch-operator to the opensearch-test namespace

https://gerrit.wikimedia.org/r/1196412

Change #1196412 merged by jenkins-bot:

[operations/deployment-charts@master] Deploy the opensearch-operator to the opensearch-test namespace

https://gerrit.wikimedia.org/r/1196412

Change #1196433 had a related patch set uploaded (by Brouberol; author: Brouberol):

[operations/deployment-charts@master] Move operator related common values away from services values and into admin_ng

https://gerrit.wikimedia.org/r/1196433

Change #1196433 merged by Brouberol:

[operations/deployment-charts@master] Move operator related common values away from services values and into admin_ng

https://gerrit.wikimedia.org/r/1196433

Change #1196497 had a related patch set uploaded (by Btullis; author: Btullis):

[operations/deployment-charts@master] Change the name of the cluster deployed in the opensearch-test namespace

https://gerrit.wikimedia.org/r/1196497

Change #1196497 merged by jenkins-bot:

[operations/deployment-charts@master] Change the name of the cluster deployed in the opensearch-test namespace

https://gerrit.wikimedia.org/r/1196497

I've loaded some test data into the cluster using a modified version of this blog post .

Specifically, I:

• Grabbed a production mapping via bking@cumin2002:~$ curl -s https://search.discovery.wmnet:9243/enwikibooks/_mapping?pretty

• Removed all the custom analyzers from the json blob via ChatGPT. Note that I only operated on the content mapping (there are also general and archive mappings in the json).

• Uploaded the mapping to the test instance curl -u ${PW} -H 'Content-Type: application/json' -XPUT $es/$index -d @mapping.json

• Followed step 3 and 4 from the blog post (split the dump, then PUT the data via OpenSearch bulk API

Verified that that data is now searchable:

Request:

curl -u ${PW}   -X GET "$es/enwikibooks/_search?pretty"   -H 'Content-Type: application/json'   -d '{
    "query": {
      "match": {
        "all": "mint leaves"
      }
    },
    "_source": ["title", "opening_text", "auxiliary_text"]
  }'

Response:

  "took" : 515,                                                                                                       [140/2315]
  "timed_out" : false,
  "_shards" : {
    "total" : 1,
    "successful" : 1,
    "skipped" : 0,
    "failed" : 0
  },
  "hits" : {
    "total" : {
      "value" : 4624,
      "relation" : "eq"
    },
    "max_score" : 18.301796,
    "hits" : [
      {
        "_index" : "enwikibooks",
        "_id" : "424726",
        "_score" : 18.301796,
        "_source" : {
          "opening_text" : "Cookbook | Recipes | Ingredients | Equipment | Techniques | Cookbook Disambiguation Pages | Recipes
There are two ways to make this mint infusion. You can either use dried up mint leaves or fresh mint leaves.",
          "auxiliary_text" : [
            "Mint Tea Category Beverage recipes Time 3–10 minutes Difficulty"
          ],
...

One thing I also noticed is that the audit index is already very large:

curl -u ${PW} -s ${es}/_cat/indices
green open security-auditlog-2025.10.21 o7CJelhPTyGRRHLC3RF0Sw 1 1    545   0   1.5gb 779.5mb

It also seemed to get quite a bit larger with every API call (I was running a for loop against 500 chunks of index). We don't want to max out our disk space due to audit logging, so we need to figure out a plan.