Kodo-X Split¶

Note

Please read this document carefully. Ensure that backup measures and rollback plans are in place during implementation to guarantee system security and stability.

Overview¶

When processing a large number of write requests, the resource consumption of kodo-x may be concentrated on specific types of processing tasks, especially metric data processing. This can cause other data to be consumed in a timely manner, leading to data accumulation.

Solution:

Create a standalone kodo-x instance dedicated to processing metric data, reducing resource consumption on the existing instance.
Modify the existing kodo-x instance so that it no longer processes metric data, focusing on other tasks.

Solution Architecture Diagram¶

Before the split:

After the split:

Prerequisites¶

You have operation permissions for the Guance cluster.
Plan the scheduling resources for kodo-x and the split kodo-x-metric in advance.

Implementation Plan¶

1. Service Backup¶

kubectl get deploy -n forethought-kodo  kodo-x -o yaml > kodo-x-deploy.yaml

2. Create the kodo-x-metric Service¶

2.1 Create the kodo-x-metric ConfigMap¶

kubectl get cm -n forethought-kodo kodo-x -o yaml > kodo-x-metric-cm.yaml

Modify the following information:

Change name to kodo-x-metric.
Delete the resourceVersion and uid fields.

kubectl apply -f kodo-x-metric-cm.yaml

2.2 Create the kodo-x-metric Deployment¶

kubectl get deploy -n forethought-kodo  kodo-x -o yaml  > kodo-x-deploy.yaml

Modify the following information:

Change metadata.name to kodo-x-metric.
Delete resourceVersion and uid from metadata.
Change the name under volumeMounts and volumes to kodo-x-metric.

...
metadata:
  name: kodo-x-metric    ##### Change to kodo-x-metric
  namespace: forethought-kodo 
  resourceVersion: "129925433"   ##### Delete
  uid: e4f0c541-4d91-47d4-82c8-91eec2757cb2   ##### Delete
...
        volumeMounts:
        - mountPath: /kodo/config/config.yaml
          name: kodo-x-metric  ##### Change to kodo-x-metric
          subPath: config.yaml

      volumes:
      - configMap:
          defaultMode: 420
          name: kodo-x-metric ##### Change to kodo-x-metric
          optional: false
        name: kodo-x-metric ##### Change to kodo-x-metric

kubectl apply -f kodo-x-deploy.yaml

3. Create the kodo-x-metric Service¶

3.1 Modify kodo-x-metric Parameters¶

Configuration file: kodo-x-metric.yaml (other configurations remain the same as the existing kodo-x.yaml)

global:
    metric_workers: 32  # Increase metric data processing capacity
    object_workers: 8
    keyevent_workers: 8
    workers: 0  # Disable the default worker
    log_workers: 0  # Disable log processing
    backup_log_workers: 0  # Disable backup log processing

Refer to the Application Service Configuration Guide for more information.

Restart the service:

kubectl rollout restart -n forethought-kodo deploy kodo-x-metric

3.2 Modify kodo-x Parameters¶

The existing kodo-x instance will only handle non-metric tasks, further optimizing resource allocation.

Configuration file: kodo-x.yaml

global:
    metric_workers: -1  # Disable metric data processing
    object_workers: -1  # Disable object processing
    keyevent_workers: -1  # Disable key event processing