Category: HA / DR

Yeva Byzek walks us through a disaster recovery scenario when running Apache Kafka:

Imagine:

Disaster strikes—catastrophic hardware failure, software failure, power outage, denial of service attack or some other event causes one datacenter with an Apache Kafka^® cluster to completely fail. Yet Kafka continues running in another datacenter, and it already has a copy of the data from the original datacenter, replicated to and from the same topic names. Client applications switch from the failed cluster to the running cluster and automatically resume data consumption in the new datacenter based on where it left off in the original datacenter. The business has minimized downtime and data loss resulting from the disaster, and continues to run its mission critical applications.

Ultimately, enabling the business to continue running is what disaster recovery planning is all about, as datacenter downtime and data loss can result in businesses losing revenue or entirely halting operations. To minimize the downtime and data loss resulting from a disaster, enterprises should create business continuity plans and disaster recovery strategies.

Distributed data sources can still succumb to disaster and many of the same policies that people learn when working with relational databases apply to things like Kafka as well.

Eva Nahari explains different techniques to back up Cloudera Search data, as well as setting up disaster recovery:

If you have the raw data in HDFS (which most do, and which you should!), the most straightforward way to have a hot-warm disaster recovery setup is to use our Backup and Disaster Recovery tool. It allows you to set up regular incremental updates between two clusters. You then have the option of using MapReduce Indexer or Spark Indexer to regularly index the raw data in your recovery cluster and append to a running Solr service in that same recovery cluster. This way you can easily switch over from one Solr service to the backup Solr service if you experience downtime in the original cluster.

The lag would be depending on the network between the clusters and how frequent you transfer data between the clusters. To some extent it would also depend on how long time you need (i.e. how much resources you have available) to complete the MapReduce or Spark indexing workload and append it (using the Cloudera Search GoLive feature) into Solr active indexes on the recovery site.

Read on for several options.

Ben Summer reviews a Kafka catstrophe and explains how to avoid it:

Here at New Relic, the Edge team is responsible for the pipelines that handle all the data coming into our company. We were an early adopter of Apache Kafka, which we began using to power this data pipeline. Our initial results were outstanding. Our cluster handled any amount of data we threw at it; it showed incredible fault tolerance and scaled horizontally. Our implementation was so stable for so long that we basically forgot about it. Which is to say, we totally neglected it. And then one day we experienced a catastrophic incident.

Our main cluster seized up. All graphs, charts, and dashboards went blank. Suddenly we were totally in the dark — and so were our customers. The incident lasted almost four hours, and in the end, an unsatisfactory number of customers experienced some kind of data loss. It was an epic disaster. Our Kafka infrastructure had been running like a champ for more than a year and suddenly it had ground to a halt.

This happened several years ago, but to this day we still refer to the incident as the “Kafkapocalypse.”

Ben has a couple interesting stories and some good rules of thumb for maintaining a Kafka cluster.