Configure Multiple Nodepool Implementations

Lucidworks recommends isolating search workloads from analytics workloads using multiple node pools. The included scripts do not do this for you; this is a manual process.

See Workload Isolation with Multiple Node Pools for more information.

Run the customize_fusion_values.sh script with the following arguments:

./customize_fusion_values.sh \
-c demo \ (1)
-n fusion-lw \ (2)
--provider k8s \
--node-pool {} \ (3)
--with-resource-limits \
--with-affinity-rules \
--with-replicas

1 The name of your cluster. This example uses demo.

2 The name of your namespace.

By using the empty value {} as the value for the --node-pool argument, the script leaves the nodePools values unassigned. Assign these values in a later step.

The example above outputs the following:

Created Fusion custom values yaml: k8s_ demo_fusion-lw_fusion_values.yaml
Created Monitoring custom values yaml: k8s_ demo_fusion-lw_monitoring_values.yaml. Keep this file handy as you'll need it to customize your Monitoring installation.
Create k8s_ demo_fusion-lw_upgrade_fusion.sh for upgrading you Fusion cluster. Please keep this script along with your custom values yaml file(s) in version control.
Update nodeSelector values for k8s_ demo_fusion-lw_fusion_values.yaml and update the solr to use nodepool

Update Solr in k8s_demo_fusion-lw_fusion_values.yaml to add nodePools. The fusion_node_type value connects the Solr nodePools values to the corresponding node pool.

solr:
  nodePools:
    - name: "" (1)
    - name: "analytics" (2)
      javaMem: "-Xmx6g"
      replicaCount: 6
      storageSize: "100Gi"
      nodeSelector:
        fusion_node_type: analytics (3)
      resources:
        requests:
          cpu: 2
          memory: 12Gi
        limits:
          cpu: 3
          memory: 12Gi
    - name: "search" (4)
      javaMem: "-Xms11g -Xmx11g"
      replicaCount: 12
      storageSize: "50Gi"
      nodeSelector:
        fusion_node_type: search (5)
      resources:
        limits:
          cpu: "7700m"
          memory: "26Gi"
        requests:
          cpu: "7000m"
          memory: "25Gi"
  nodeSelector:
    cloud.google.com/gke-nodepool: default-pool (6)
...

1 The empty string "" is the suffix for the default partition.

2 Overrides the settings for the analytics Solr pods.

3 Assigns the analytics Solr pods to the node pool and attaches the label fusion_node_type=analytics. You can use the fusion_node_type property in Solr auto-scaling policies to govern replica placement during collection creation.

4 Overrides the settings for the search Solr pods

5 Assigns the search Solr pods to the node pool and attaches the label fusion_node_type=search.

Sets the default settings for all Solr pods, if not specifically overridden in the nodePools section above.

Do not edit the nodePools value "".

Only use a single fusion_node_type value per Solr nodeSelector value.

Update k8s_demo_fusion-lw_fusion_values.yaml to add appropriate nodeSelector values for all desired services.

Run the upgrade_fusion.sh script:

./k8s_demo_fusion-lw_upgrade_fusion.sh

The output resembles the following:

namespace/fusion-lw created

Created namespace fusion-lw with owner label

Hang tight while we grab the latest from your chart repositories...

... (1)

NAME: fusion-lw
LAST DEPLOYED: Wed Sep 15 18:25:55 2021
NAMESPACE: fusion-lw
STATUS: deployed
REVISION: 1

...

NAME            NAMESPACE       REVISION        UPDATED                                 STATUS          CHART           APP VERSION
fusion-lw       fusion-lw       1               2021-09-15 18:25:55.1470127 -0500 CDT   deployed        fusion-5.4.2    5.4.2

1	The `…` indicates some parts of the output are removed to shorten the example response.

Install Prometheus and Grafana so you can monitor Fusion metrics by running the following:

./install_prom.sh --provider k8s -c demo -n fusion-lw

The output resembles the following:

Adding the stable chart repo to helm repo list
"stable" has been added to your repositories

Installing Prometheus and Grafana for monitoring Fusion metrics ... this can take a few minutes.

...

NAME: fusion-lw-monitoring
LAST DEPLOYED: Wed Sep 15 22:59:28 2021
NAMESPACE: fusion-lw
STATUS: deployed
REVISION: 1
NOTES:
The Prometheus server can be accessed via port 80 on the following DNS name from within your cluster:
fusion-lw-monitoring-prometheus-server.fusion-lw.svc.cluster.local


...

Successfully installed Prometheus and Grafana into the fusion-lw namespace.

NAME                    NAMESPACE       REVISION        UPDATED                                 STATUS          CHART                   APP VERSION
fusion-lw               fusion-lw       1               2021-09-15 18:25:55.1470127 -0500 CDT   deployed        fusion-5.4.2            5.4.2
fusion-lw-monitoring    fusion-lw       1               2021-09-15 22:59:28.9054416 -0500 CDT   deployed        fusion-monitoring-1.0.0 1.0.0

Update the Solr collections to use specific nodes. Apply the Solr policy and run update_app_coll_layout.sh. This script deletes the collections and recreates them with the new policy. Check the scripts for details on collection names.