ICAgent Constraints

CCE cluster type

Currently, ICAgent can be installed only in the following CCE cluster types:

• CCE standard clusters
• CCE Turbo clusters

ICAgent cannot be installed in the following CCE cluster types:

• CCE Autopilot clusters
• CCE Agile clusters

N/A

Self-built Kubernetes cluster scale

When installing ICAgent in a self-built Kubernetes cluster, ensure the cluster meets these requirements:

• Maximum nodes: 50
• Maximum pods: 400

N/A

File encoding

UTF-8 and GBK are supported. Other encoding formats may cause garbled characters.

For example, binary files.

N/A

Log file size

No restrictions.

N/A

Log file rotation

ICAgent supports configuration of fixed log file names or fuzzy match of log file names. You need to rotate log files manually.

N/A

Log collection path

Linux

• Collection paths support recursion. You can use double asterisks (**) to collect logs from up to five directory levels. Example: /var/logs/**/a.log
• Collection paths support fuzzy match. You can use an asterisk (*) to represent one or more characters of a directory or file name. Example: /var/logs/*/a.log or /var/logs/service/a*.log
• If the collection path is set to a directory, for example, /var/logs/, only .log, .trace, and .out files in the directory are collected. If the collection path is set to name of a text file, that file is directly collected.
• After you enable Allow Repeated File Collection, one host log file can be collected to multiple log streams. After you disable it, each collection path must be unique. That is, the same path of the same host cannot be configured for different log groups and log streams.

Windows

• Collection paths support recursion. You can use double asterisks (**) to collect logs from up to five directory levels. Example: C:\var\service\**\a.log
• Collection paths support fuzzy match. You can use an asterisk (*) to represent one or more characters of a directory or file name. Examples: C:\var\service\*\a.log and C:\var\service\a*.log
• Each collection path must be unique. That is, the same path of the same host cannot be configured for different log groups and log streams.
• Each collection path must be unique. That is, the same path of the same host cannot be configured for different log groups and log streams.

N/A

Symbolic link

Symbolic links are not supported.

N/A

Single log size

The default size of each log is 500 KB. The maximum size is 1,024 KB. Excess content will be truncated by ICAgent.

N/A

Regular expression

ICAgent supports only RE2 regular expressions. For details, see Syntax.

N/A

Repeated file collection

The Allow Repeated File Collection option is available only for Linux ECSs, not for Windows ECSs.

After you enable it, one host log file can be collected to multiple log streams. After you disable it, each collection path must be unique. That is, the same log file in the same host can be collected to only one log stream.

N/A

File opening

Files are opened when being read, and closed after being read.

N/A

First log collection

All logs are collected.

N/A

Container logs

In container scenarios, only the following storage volume types are supported for volume mounting directories:

• hostPath volumes, with an empty mount subpath
• emptyDir volumes, with an empty mount subpath

Container scenarios

Log collection rate

Raw logs of a single node are collected at a rate up to 50 MB/s.

Service quality cannot be ensured if this limit is exceeded.

Monitored directories

The default directory recursion depth is 5, and the maximum is 20, with up to 500 files.

N/A

Monitored files

Container scenarios

• ICAgent can collect a maximum of 20 log files from a volume mounting directory.
• ICAgent can collect a maximum of 1,000 standard container output log files. These files must be in JSON format.

VM scenarios: A maximum of 500 files are supported.

In container scenarios, only the following storage volume types are supported for volume mounting directories:

• hostPath volumes, with an empty mount subpath
• emptyDir volumes, with an empty mount subpath

Default resource constraints

CPU:

• If the ICAgent version is earlier than 5.12.200, up to two CPU cores are consumed.
• If the ICAgent version is 5.12.200 or later, up to two CPU cores are consumed when there are no more than 4 node cores, and the maximum CPU resources consumed are log2(number of node cores) when there are more than 4 node cores.

Memory:

The maximum memory consumed is min{4 GB, Physical memory/2}. A restart is triggered if this memory limit is exceeded. "min{4 GB, Physical memory/2}" means that the smaller value between half of the physical memory and 4 GB is used.

N/A

Resource limit reached

A forcible restart is triggered. Logs may be lost or duplicate if rotated during the restart.

N/A

Agent installation, upgrade, or uninstallation

No restrictions.

N/A

Configuration update

Configuration updates take effect in 1 to 3 minutes.

N/A

Dynamic configuration loading

Console configurations can be dynamically delivered. The update of one configuration does not affect other configurations.

N/A

Configurations

No restrictions.

N/A

Tenant isolation

Tenants are isolated from each other by default.

N/A

Log collection delay

Normally, the delay from writing logs to the disk to collecting the logs is less than 10s (congestion not considered).

N/A

Log upload

File changes are read and uploaded immediately once detected. One or more logs can be uploaded a time.

N/A

Network error handling

Network exceptions trigger retries at an interval of 5s.

N/A

Resource quota used up

If the resources allocated to the ICAgent are insufficient due to massive amounts of logs, the ICAgent continues and retries upon a failure. Logs will be stacked if resources are still insufficient.

N/A

Max. retry timeout

Retry attempts are periodically made.

N/A

Status check

The collector status is monitored through heartbeat detection.

N/A

Checkpoint timeout

Checkpoints are automatically deleted if no updates are made within 12 hours.

N/A

Checkpoint saving

Checkpoints are updated if logs are reported successfully.

N/A

Checkpoint saving path

By default, checkpoints are saved in /var/share/oss/manager/ICProbeAgent/internal/TRACE.

N/A

Logs lost or repeated

ICAgent uses multiple mechanisms to ensure log collection reliability and prevent data loss. However, logs may be lost or repeated in the following scenarios:

• The log rotation policy of CCE is not used.
• Log files are rotated at a high speed, for example, once per second.
• If the rsyslog service does not exist or is disabled, ICAgent cannot collect system logs.
• The container running time, for example, shorter than 30s, is extremely short.
• A single node generates logs at a high speed, exceeding the allowed transmit bandwidth or log collection speed. It is recommended that the log generation speed of a single node be lower than 50 MB/s.

When ICAgent is restarted, identical data may be collected around the restart time.

N/A

OpenStack

http://169.254.169.254/openstack/latest/meta_data.json

Obtain the metadata, name, and ID of a node.

http://169.254.169.254/openstack/latest/securitykey

Obtain a temporary AK/SK and security token with an agency.

http://169.254.169.254/latest/meta-data/public-ipv4

Obtain the EIP bound to a node.

CCE

http://127.0.0.1:4194/api/v2.0/ps

Obtain process information with the cAdvisor API.

http://127.0.0.1:4194/api/v1.2/docker

Obtain all container metrics with the cAdvisor API.

http://nodeip:10255/pods

Obtain pod information with a Kubernetes API.

#icmgr-service 
{podlb}:30200

ICAgent registration

icmgr-controller
{podlb}:30201

ICAgent status configuration

#als-access 
{podlb}:8102

Log reporting

#ams-access 
{podlb}:8149

Metric reporting

#ats-access apm
{podlb}:8923

Data reporting to APM