Diskprediction Module

The diskprediction module supports two modes: cloud mode and local mode. In cloud mode, the disk and Ceph operating status information is collected from Ceph cluster and sent to a cloud-based DiskPrediction server over the Internet. DiskPrediction server analyzes the data and provides the analytics and prediction results of performance and disk health states for Ceph clusters.

Local mode doesn’t require any external server for data analysis and output results. In local mode, the diskprediction module uses an internal predictor module for disk prediction service, and then returns the disk prediction result to the Ceph system.

Local predictor: 70% accuracy

Cloud predictor for free: 95% accuracy

Enabling

Run the following command to enable the diskprediction module in the Cephenvironment:

ceph mgr module enable diskprediction_cloud
ceph mgr module enable diskprediction_local

Select the prediction mode:

ceph config set global device_failure_prediction_mode local

or:

ceph config set global device_failure_prediction_mode cloud

To disable prediction,:

ceph config set global device_failure_prediction_mode none

Connection settings

The connection settings are used for connection between Ceph and DiskPrediction server.

Local Mode

The diskprediction module leverages Ceph device health check to collect disk health metrics and uses internal predictor module to produce the disk failure prediction and returns back to Ceph. Thus, no connection settings are required in local mode. The local predictor module requires at least six datasets of device health metrics to implement the prediction.

Run the following command to use local predictor predict device life expectancy.

ceph device predict-life-expectancy <device id>

Cloud Mode

The user registration is required in cloud mode. The users have to sign up their accounts at https://www.diskprophet.com/#/ to receive the following DiskPrediction server information for connection settings.

Certificate file path: After user registration is confirmed, the system will send a confirmation email including a certificate file download link. Download the certificate file and save it to the Ceph system. Run the following command to verify the file. Without certificate file verification, the connection settings cannot be completed.

DiskPrediction server: The DiskPrediction server name. It could be an IP address if required.

Connection account: An account name used to set up the connection between Ceph and DiskPrediction server

Connection password: The password used to set up the connection between Ceph and DiskPrediction server

Run the following command to complete connection setup.

ceph device set-cloud-prediction-config <diskprediction_server> <connection_account> <connection_password> <certificate file path>

You can use the following command to display the connection settings:

ceph device show-prediction-config

Additional optional configuration settings are the following:

• diskprediction_upload_metrics_interval

• Indicate the frequency to send Ceph performance metrics to DiskPrediction server regularly at times. Default is 10 minutes.

• diskprediction_upload_smart_interval

• Indicate the frequency to send Ceph physical device info to DiskPrediction server regularly at times. Default is 12 hours.

• diskprediction_retrieve_prediction_interval

• Indicate Ceph that retrieves physical device prediction data from DiskPrediction server regularly at times. Default is 12 hours.

Diskprediction Data

The diskprediction module actively sends/retrieves the following data to/from DiskPrediction server.

Metrics Data

• Ceph cluster status

• Ceph mon/osd performance counts

Mon:

Osd:

• Ceph pool statistics

• Ceph physical device metadata

• Ceph each objects correlation information

• The module agent information

• The module agent cluster information

• The module agent host information

SMART Data

• Ceph physical device SMART data (provided by Ceph devicehealth module)

Prediction Data

• Ceph physical device prediction data

Receiving predicted health status from a Ceph OSD disk drive

You can receive predicted health status from Ceph OSD disk drive by using thefollowing command.

ceph device get-predicted-status <device id>

The get-predicted-status command returns:

{
    "near_failure": "Good",
    "disk_wwn": "5000011111111111",
    "serial_number": "111111111",
    "predicted": "2018-05-30 18:33:12",
    "attachment": "sdb"
}

The near_failure attribute for disk failure prediction state indicates disk life expectancy in the following table.

Debugging

If you want to debug the DiskPrediction module mapping to Ceph logging level,use the following command.

[mgr]
 
    debug mgr = 20

With logging set to debug for the manager the module will print out loggingmessage with prefix mgr[diskprediction] for easy filtering.