This guide provides instructions on how to use the logging infrastructure for DOCA services on top of NVIDIA® BlueField® DPU.
Fluent Bit is a fast log collector that collects information from multiple sources and then forwards the data onward using Fluent.
On NVIDIA DPUs, the Fluent Bit logger can be easily configured to collect system data and the logs from the different DOCA services.
The deployment is based on a recommended configuration template for the existing Fluent Bit container.
For information about the deployment of DOCA containers on top of the BlueField DPU, refer to NVIDIA DOCA Container Deployment Guide.
The following is an example YAML file for deploying the Fluent Bit pod:
apiVersion: v1
kind: Pod
metadata:
name: fluent-bit
spec:
hostNetwork: true
containers:
- name: fluent-bit
image: fluent/fluent-bit:latest
imagePullPolicy: Always
# Example resource definitions
resources:
requests:
memory: "100Mi"
cpu: "200m"
limits:
memory: "200Mi"
cpu: "300m"
volumeMounts:
- name: varlog
mountPath: /var/log
- name: config-file
mountPath: /fluent-bit/etc/fluent-bit.conf
volumes:
- name: varlog
hostPath:
path: /var/log
- name: config-file
hostPath:
path: /opt/mellanox/doca/services/fluent-bit.conf
type: File
As explained in the "Configuration" section, Fluent Bit uses a configuration file. As such, to ensure that the example YAML file is shared from the DPU to the deployed Fluent Bit container, use the following:
path: /opt/mellanox/doca/services/fluent-bit.conf
The path below is just an example for where the user can place the fluent-bit.conf file. The file could be placed in a different directory on the DPU as long as the YAML file points to the updated location.
The Fluent Bit configuration file should have the following sections:
[SERVICE] – to define the service specifications
[INPU] – to define folders to collect logs from (there could be multiple inputs)
[OUTPUT] – IP and port to stream the data to
Example configuration file:
[SERVICE]
Flush 2
Log_Level info
Daemon off
Parsers_File parsers.conf
HTTP_Server On
HTTP_Listen 0.0.0.0
HTTP_Port 2020
[INPUT]
Name tail
Tag kube.*
Path /var/log/containers/*.log
Parser docker
Mem_Buf_Limit 5MB
Skip_Long_Lines On
Refresh_Interval 10
[INPUT]
Name tail
Tag sys.*
Path /var/log/doca/*/*.log
Mem_Buf_Limit 5MB
Skip_Long_Lines On
Refresh_Interval 10
[OUTPUT]
Name es
Match *
Host 10.20.30.40
Port 9201
Index fluent_bit
Type cpu_metrics
The most important field to pay attention to is Path for the INPUT section. DOCA services report their logs to a unique directory under /var/log/doca/<service_name>/*.log per the respective DOCA service. As such, the configuration above defines the /var/log/doca/*/*.log input definition.
More information about the full specifications can be found in the official Fluent Bit manual.
For container-related troubleshooting, refer to the "Troubleshooting" section in the NVIDIA DOCA Container Deployment Guide.
For general troubleshooting, refer to the NVIDIA DOCA Troubleshooting Guide.
When copying the above YAML file, it is possible that the container infrastructure logs give an error related to RFC 1123". These errors are usually a result of a spacing error in the file, which sometimes occur when copying the file as is from this page. To fix this issue, make sure that only the space character (' ') is used as a spacer in the file and not other whitespace characters that might have been added during the copy operation.