Is endpoint cpm accurate?

[JHub-Wei]

Please answer these questions before submitting your issue.

• Why do you submit this issue?
• [ * ] Question or discussion
• Bug
• Requirement
• Feature or performance improvement

---

Question

• What do you want to know?
  We perform a pressure test on an endpoint at 1000 TPS. However, the CPM value of the endpoint is only 7438 on the dashboard during this period. 7438 / 60 = 124,it doesn't match the actual tps.Why is this happening?

The test service is a single instance. The OAP cluster is in cluster mode and ZooKeeper is used.

---

Bug

• Which version of SkyWalking, OS, and JRE?
  8.4.0，jdk1.8

• Which company or project?

• What happened?
  If possible, provide a way to reproduce the error. e.g. demo application, component version.

---

Requirement or improvement

• Please describe your requirements or improvement suggestions.

[wu-sheng]

If OAP is running in the cluster mode with ZooKeeper, you need to make sure their binding-IPs are not 0.0.0.0. Otherwise, you are going to face dirty write easily.

[wu-sheng]

Quote answer from #7484 (reply in thread)

---

It depends. Could be SW_CORE_GRPC_HOST, or internalComHost or internalComPort, read document before you take actions, https://skywalking.apache.org/docs/main/latest/en/setup/backend/backend-cluster/#zookeeper-coordinator

---

[JHub-Wei]

Of course, their binding-IPs are not 0.0.0.0.The environment variable SW_CLUSTER_ZK_HOST_PORT is configured.

[wu-sheng]

SW_CLUSTER_ZK_HOST_PORT is for ZooKeeper server address. I am asking for OAP binding-IPs.

[JHub-Wei]

What Does OAP Binding-IPs Mean?Is the environment variable SW_CORE_GRPC_HOST?

[wu-sheng]

It depends. Could be SW_CORE_GRPC_HOST, or internalComHost or internalComPort, read document before you take actions, https://skywalking.apache.org/docs/main/latest/en/setup/backend/backend-cluster/#zookeeper-coordinator

[JHub-Wei]

Thx，I'll take a look first.

[JHub-Wei]

If the OAP is not in cluster mode before, do I need to restart the service integrated with the agent after the OAP is changed to cluster mode?

[wu-sheng]

No, you don't need to reboot agent.

[larrying]

did you resolve this problem? i faced the same problem

[JHub-Wei]

Yes, We set SW_CORE_GRPC_HOST in deployment, and it works.
- name: SW_CORE_GRPC_HOST
valueFrom:
fieldRef:
fieldPath: status.podIP

[larrying]

thank you, it works