Skip to main content

Configuration

eBPF Agent Meltdown​

Meltdown Principle​

The eBPF Agent meltdown strategy determines whether to perform Agent meltdown based on the Node's host CPU usage and available memory size.

  • The default CPU threshold is 75%~85%. When the Node's CPU usage exceeds 85%, Agent meltdown is triggered.

  • The default available memory threshold is 1GB~2GB. When the Node's available memory is less than 1GB, Agent meltdown is triggered.

  • After Agent meltdown, when the host CPU usage drops below 75% and available memory exceeds 2GB, the Agent resumes normal operation.

Note: When meltdown occurs, the eBPF Agent process will exit. After meltdown recovery, the eBPF Agent process will restart.

Configuration Modification​

Modify ebpf_cpu_blow_threshold and ebpf_mem_blow_threshold in tingyun-common-config in tingyunagent.yaml.

apiVersion: v1
kind: ConfigMap
metadata:
  name: tingyun-common-config
  namespace: tingyun
data:
  VERSION: |
    2.4.1.0
  tingyun-common.yaml: |
    # Whether to enable the EBPF network data collection function, true or false
    # If you enable the EBPF function, you also need to modify the Daemonset tingyun-linux-agent requests cpu to 0.5, memory to 1Gi, and restart the Daemonset tingyun-linux-agent
    ebpf_enabled: false
    
    # EBPF network module CPU meltdown threshold: enable threshold, disable threshold
    # Unit: percent
    # When the Node's CPU usage exceeds the disable threshold, the EBPF network module switches to meltdown state
    # When in meltdown state, and CPU usage drops below the enable threshold and available memory exceeds the enable threshold, the EBPF network module is enabled
    ebpf_cpu_blow_threshold: 75,85
    
    # EBPF network module available memory meltdown threshold: disable threshold, enable threshold
    # Unit: bytes
    # When the Node's available memory is less than the disable threshold, the EBPF network module switches to meltdown state
    # When in meltdown state, and CPU usage drops below the enable threshold and available memory exceeds the enable threshold, the EBPF network module is enabled
    ebpf_mem_blow_threshold: 1073741824,2147483648

Then update using the yaml file.

kubectl apply -f tingyunagent.yaml

If the tingyunagent.yaml installation file is lost, you can use the edit configmap command to modify the configuration.

kubectl edit configmap tingyun-common-config -n tingyun

View meltdown logs:

kubectl -n tingyun exec -ti tingyun-linux-agent-vrtwk -- cat /opt/agent/ebpf/restart.log

Sample output:

ebpf agent stop, because of feature disable                  # ebpf_enabled is false
ebpf agent start, because of crash restart or blow recover   # ebpf_enabled is true, Agent started
ebpf agent stop, because of trigger cpu blow                 # CPU high triggers meltdown
ebpf agent start, because of crash restart or blow recover   # Meltdown recovery, Agent started
ebpf agent stop, because of trigger memory blow              # Insufficient available memory triggers meltdown
ebpf agent start, because of crash restart or blow recover   # Meltdown recovery, Agent started
ebpf agent not start because of memory blow                  # Insufficient available memory, Agent not started

Log Level and Audit Mode​

Modify the configuration items log.level and log.audit in the ebpf.conf file in tingyun-sys-config in tingyunagent.yaml.

apiVersion: v1
kind: ConfigMap
metadata:
  name: tingyun-sys-config
  namespace: tingyun
data:
  ebpf.conf: |
    #
    # Config item: log.filecount
    # Purpose: Number of log files to retain
    # Default: 10
    # Requires pod restart after modification
    log.filecount=3
    # 
    # Config item: log.filesize
    # Purpose: Size of a single log file
    # Default: 100
    # Unit: MB
    # Requires pod restart after modification
    log.filesize=100
    # 
    # Config item:  log.level
    # Purpose:    Log data filter level
    # Possible values: debug, info, warning, error
    # Default:  info
    # Requires pod restart after modification
    log.level=info
    #
    # Config item:  log.audit
    # Purpose:    Enable audit mode logs
    # Possible values: true, false
    # Default:  false
    # Requires pod restart after modification
    log.audit=false

Then update using the yaml file.

kubectl apply -f tingyunagent.yaml

If the tingyunagent.yaml installation file is lost, you can use the edit configmap command to modify the configuration.

kubectl edit configmap tingyun-sys-config -n tingyun

Then delete the existing tingyun-linux-agent pod to let the pod be recreated.

kubectl -n tingyun delete pod tingyun-linux-agent-ddefb

View work logs:

kubectl -n tingyun exec -ti tingyun-linux-agent-vrtwk -- cat /opt/agent/ebpf/logs/ebpf-agent.log