FAQ
This document was translated by GPT-4
# 1. Deployment
What's the difference between all-in-one deployment mode and standard deployment mode?
Answer: All-in-one deployment means the storage components
clickhouse
andmysql
aren't assigned corresponding PVCs, but usehostPath
deployment mode instead. If your Kubernetes cluster has multiple nodes, restarting thedeepflow-clickhouse/mysql
Pod might cause it to drift to other nodes, making the data previously collected unsearchable. It is recommended to use all-in-one deployment for trials or testing/POC stages.How long is data generally retained, and can this be adjusted?
Answer: Different types of data have different retention periods. You can refer to the server.yaml (opens new window) file to check the retention period of each data type. The retention time can be adjusted
before the first deployment
. Adjust the default settings and complete the installation when installing via helm.How can I use an external MySQL/Clickhouse?
Answer: Please refer to the sections on utilizing a managed MySQL solution and utilizing a managed ClickHouse solution in the production environment deployment recommendations.
The deployment specifications include
mysql
andclickhouse
storage components. What's the difference between them?Answer:
mysql
stores metadata information obtained from the deployment cluster, such as virtual machines, Kubernetes resources, sync collector information, etc.clickhouse
stores real-time collected data, such as network stream logs from the cluster, and performs aggregation analysis.Why is there no data on Grafana after deployment?
Answer: Please troubleshoot using the following steps:
Check if all Pods are running normally: Run the
kubectl get pods -n deepflow
command and confirm all Pods are inRunning
status.Check whether DeepFlow Agent and DeepFlow Server have been successfully connected. Use
deepflow-ctl domain list
to check if the service domain has been successfully created, then usedeepflow-ctl agent list
to check if theSTATE
is inNORMAL
status.If the
Network - X
dashboards have no data, check whether the network card name complies with the capture rules. Use thedeepflow-ctl agent-group-config example | grep tap_interface_regex
command to check the default capture scope. If you've used a custom CNI or constructed the network in other ways, add the network card match rule totap_interface_regex
, and then complete the changes by updating the agent configuration.If only the
Application - X
dashboards have no data, ensure that the application protocol used within your cluster meets the support list requirements.
I've set up OpenTelemetry data integration/I'd like to use eBPF tracing and network tracing capabilities offered by DeepFlow, but I can't see any data on the
Distributed Tracing
dashboard.Answer: Please troubleshoot using the following steps:
When using OpenTelemetry integration:
Ensure that the application has integrated with the OTel SDK or started the OTel Agent.
Confirm you've completed the configuration according to the DeepFlow configuration. On the container node where
deepflow-agent
is located, use thenetstat -alntp | grep 38086
command to check if this function has started normally. If the configuration is complete, check theNetwork - Flow Log
for flow logs withServer Port
listed as 38086.Check the
Application - K8s Pod Map
dashboard to see if there's traffic from the application to the otel-agent container node. Ensure this network link is unobstructed and actually processing requests.In the
Application - Request Log
dashboard, verify that the outgoing requests aren't abnormal.
When using eBPF capabilities:
Confirm that the server kernel version meets requirements.
Check all instances of
deepflow-agent
: Use thekubectl logs -n deepflow ds/deepflow-agent | grep 'ebpf collector'
command to check if the eBPF module has started normally. Use thekubectl logs -n deepflow ds/deepflow-agent | grep TRACER
command to ensure eBPF Tracer is functioning correctly.
# 2. Product
After installing and deploying, what should I do next? Are there product cases or usage scenarios you can share?
Answer: You can see our shared cases in our Start Observability (opens new window) and troubleshooting (opens new window) blog series. Also, you can revisit previous shares on our Bilibili account (opens new window).
I think some features aren't optimal and would like to offer suggestions. How can I do this?
Answer: You're welcome to submit Feature Requests on Github Issue (opens new window). If you already have a mature idea, you can put it into practice and submit it directly at GithubPR (opens new window).
Where can I keep track of the latest developments of DeepFlow?
Answer: You can check our latest release overview in the release notes or follow our latest blog (opens new window).
# 3. Contact Us
If the above help cannot resolve your problem, you can submit your questions via Github Issue (opens new window) or contact us (opens new window) directly for discussion.