View Only

AIOps v3.1 - Series 05 : Training Logs, Events and Incidents

By Jeya Gandhi Rajan M posted Wed May 12, 2021 05:14 PM


This article explains about how to do Training for the following in Watson AIOps.

  • Log Anomaly detection
  • Event Grouping
  • Similar Incidents

1. Training - Log Anomaly Detection

This section explains about how to do Training of Log Anomaly Detection in Watson AIOps.


Here is the application page looks like.

And assume that app is accessible via link

Generate Load

The below script can be used for generating load.

Run this script and Generate load.


while (true)
    ab -n 100 -c 5

Load Live logs for Training

Enable the live logs for the training.

Keep in this state for 10 minutes.

Stop Live logs

After 10 minutes, live logs can be disbled.

Stop Load

Click on Ctrl+C to stop load script.

Start Training

Training is started.

If you have this Needs improvement status then you need to load some more logs.

Need around 10K lines logs for each component/micro service of the app.

Now the trianing status is Good and Deployed.

The trained version is v8.

2. Training - Event Grouping

This section explains about how to do Training of Event Grouping in Watson AIOps.


Sample Events Files for Training

The sample events data file is available here 02-noi-alerts.json.

Pushing events for Training

The script to push events into kafka topic alerts-noi-1000-1000 for training is available here

oc extract secret/strimzi-cluster-cluster-ca-cert --keys=ca.crt --to=- > ca.crt
export SASL_PASSWORD=$(oc get secret token --template={{.data.password}} | base64 --decode)
export SEC="-X security.protocol=SSL -X -X sasl.mechanisms=SCRAM-SHA-512 -X sasl.username=token -X sasl.password=$SASL_PASSWORD"
export BROKER=$(oc get routes strimzi-cluster-kafka-bootstrap -o=jsonpath='{.status.ingress[0].host}{"\n"}'):443

kafkacat $SEC -b $BROKER -P -t alerts-noi-1000-1000 -l ./02-noi-alerts.json

To know more about how to access kafka topics refer here accessing-kafka-topics

Enable Data flow for Training

Push training events to kafka topic

Run the below command to push training events to kafka topic.

sh ./files/

Disable Data flow

Start Training

3. Training - Similar Incidents

This section explains about how to do Training of Similar Incidents in Watson AIOps.

The article is based on the the following

  • RedHat OpenShift 4.6 on IBM Cloud (ROKS)
  • Watson AI-Ops 3.1.0


Sample Incidents for Training

Atleast 5 incidents should have been created in service-now, resolved with proper resolve comments and closed.

Here are the sample incidents from service now.

Enable Data flow for Training

Make sure the following

  • Data flow is on
  • Historical data for Initial AI-Training
  • Start Date should be past date
  • End Date could be current date

Start Training

Start the training

Released by :

Jeya Gandhi Rajan M
Vijaya Bhaskar R Siddareddi
Vijay Sukthankar (Squad Leader)

Hybrid-Cloud Squad
GSI Labs