Can Node_exporter and Blackbox_exporter be omitted from deployment?

Note:
This topic has been translated from a Chinese forum by GPT and might contain errors.

Original topic: Node_exporter 和 Blackbox_exporter 是否可以不部署

| username: 月明星稀

May I ask if Node_exporter and Blackbox_exporter can be omitted from deployment? If they are not needed, how should they be configured to be disabled?

| username: 像风一样的男子 | Original post link

Without this, the monitoring system will have no data, and it will be impossible to troubleshoot issues.

| username: tidb菜鸟一只 | Original post link

Cannot be undeployed.

| username: zhanggame1 | Original post link

If not, then scale down both Grafana and Prometheus.

| username: 胡杨树旁 | Original post link

I have always been confused about what these two services are used for. Are they used by Prometheus to collect information from various nodes?

| username: 像风一样的男子 | Original post link

Yes, the various metrics of the collection server and database are aggregated into Prometheus and displayed on Grafana.

| username: 小王同学Plus | Original post link

If you do not need to install the exporter, configure ignore_exporter: true under the host of all components. However, as mentioned above, without this, the monitoring system will have no data, making it impossible to troubleshoot issues.

| username: ffeenn | Original post link

Not deploying the entire cluster is like being blind in one eye. I’m really curious why there is an ignore_exporter parameter. Is it for debugging and development?

| username: cassblanca | Original post link

Is monitoring not needed?

| username: Fly-bird | Original post link

If it’s not monitored, it won’t be deployed.

| username: DBRE | Original post link

Node_exporter and Blackbox_exporter should only collect server-related metrics, while the database service metrics are collected by Prometheus configured with TiDB components as Prometheus targets.

| username: DBRE | Original post link

Is there an explanation of the ignore_exporter parameter in the official documentation? I couldn’t find it.

| username: Miracle | Original post link

I’m curious if the latency-related metrics in TiDB can still be found without deploying Blackbox_exporter.

| username: realcp1018 | Original post link

If you stop the exporter, you will know. The latency metrics of TiDB should be obtained from tidb-server:10080, so it shouldn’t have any impact.

| username: DBRE | Original post link

blackbox_exporter is an official black-box monitoring solution provided by Prometheus. It is one of the exporters that can probe the network using methods such as HTTP(S), DNS, TCP, and ICMP.

Currently supported application scenarios:

  • ICMP Testing
    • Host liveness detection
  • TCP Testing
    • Monitoring the status of business component ports
    • Defining and monitoring application layer protocols
  • HTTP Testing
    • Defining Request Header information
    • Checking Http status / Http Response Header / Http Body content
  • POST Testing
    • Interface connectivity
  • SSL Certificate Expiry Time
  • Custom Testing (Extensions)
| username: 月明星稀 | Original post link

Can the HTTP protocol service be disabled? I’m not sure if this service can enable TLS authentication, or if there are other authentication methods available.

| username: 小王同学Plus | Original post link

Looking at this PR: cluster: add ability to ignore monitor agent for instances by AstroProfundis · Pull Request #1492 · pingcap/tiup · GitHub, it seems to avoid conflicts.

For example, if you want to deploy instances from multiple clusters on a single host, the monitoring of these clusters might conflict. :thinking:

| username: 月明星稀 | Original post link

Can the HTTP protocol service be disabled? I’m not sure if this service can enable TLS authentication, or if there are other authentication methods available.

| username: Jellybean | Original post link

Almost all use cases will deploy these two monitoring metric collection components. I’m curious about the scenario on your side that requires disabling them.

| username: Kongdom | Original post link

Expressing the same curiosity, in what scenario would there be a need to disable the monitoring component? Is it because of concerns about background reporting?