SlideShare a Scribd company logo
1 of 24
Download to read offline
Monitoring Large-Scale
Spark Clusters at
Databricks
Nanxi Kang
Apr.12th
2
Who we are
Inventors of
• The largest open source project in data processing
• Unified platform for ETL, interactive query, machine learning, streaming, ...
Leader in Big Data ecosystem
3
What we do
We offer a cloud-based Spark platform
• Demo shard
4
Databricks platform
Databricks
Create
Configure
Terminate
Customer 1
Customer 2
ClusterClusterCluster
ClusterCluster
service
service
service
customer
requests
5
Requirements of monitoring
As an engineer, I want to
• Understand the state of the service
• Have customer-oriented metrics (uptime, performance regression, ...)
• Be alerted on abnormality
• Ease debugging and troubleshooting
6
Requirements of monitoring
As an engineer, I want to
• Understand the state of the service
• Have customer-oriented metrics (uptime, performance regression, ...)
• Be alerted on abnormality
• Ease debugging and troubleshooting
Technically
• Ingest metrics with low latency
• Aggregate metrics
• Scalability
7
Ingest and aggregate metrics
Prometheus Grafana
Alert Manager +
PagerDuty
Databricks
service
service
Metrics
latency{operation=”runCommand”} = 60 alert if latency{operation=”runCommand”} > 100
8
Ingest and aggregate metrics
Prometheus Grafana
Alert Manager +
PagerDuty
Databricks
service
service
Metrics
Kinesis
9
Latency
Prometheus Grafana
Alert Manager +
PagerDuty
Databricks
service
service
Metrics
Kinesis
~ 30 seconds
~ 5 seconds
refresh every 1 ~ 5 mins
~ 5 seconds
10
Recap Databricks platform
Databricks
Create
Configure
Terminate
Customer 1
Customer 2
ClusterClusterCluster
ClusterCluster
service
service
service
customer
requests
11
Unique challenges
• Heterogeneous environments
• Scalability: large volume of metrics
– ~300K/min from Databricks
– ~2M/min from Customers
Customer 1
Databricks
ClusterClusterCluster
service
service
service
Prometheus
12
Ingest metrics from customer env
Prometheus
Kinesis
Customer
13
Ingest Metrics from Customer Env
Prometheus Grafana
Alert Manager +
PagerDuty
Kinesis
Customer
Kinesis2Prom
14
What Next?
Let’s collect metrics!
Services can emit structured data about how they’re doing
• This is called whitebox monitoring
• Example: “how long does it take to finish a spark command?”
• Golden metrics (Latency, Traffic, Error, and Saturation)
15
Whitebox monitoring
During a recent AWS S3 outage in us-east-1, our services
report failures to create Spark clusters.
16
What’s missing?
• Hard to cover the end-to-end (cross-)service behaviors
• Whitebox metrics are great for causes, not symptoms
17
What’s missing?
• Hard to cover the end-to-end (cross-)service behaviors
• Whitebox metrics are great for causes, not symptoms
Understand the user’s experience
• Use a fake user to record the service’s behavior
• This is called Blackbox monitoring
• Example: a user records the time from “he hits the command run UI
button” to “he sees the results/errors from UI.”
18
Metrics in realtime
19
Our monitoring pipeline needs to be monitored too!
Meta-monitoring
Prometheus Grafana
Alert Manager +
PagerDuty
Meta-Prometheus
20
We can shard metrics and launch more Prometheus(es).
Scale out monitoring pipeline
Prometheus
Databricks
Prometheus
Kinesis2Prom
metric filters
21
We can shard metrics and launch more Prometheus(es).
However, we cannot join metrics across partitions.
Scale out monitoring pipeline
22
We can shard metrics and launch more Prometheus(es).
However, we cannot join metrics across partitions.
We are starting to offload extremely high-volume metrics to
Spark’s structured streaming.
Scale out monitoring pipeline
23
Monitoring is useful!
- Since rollout, all teams at Databricks started using
dashboards and alerts
Add metrics as you build features
Building scalable and robust monitoring systems is hard
Take-aways
Thanks & We Are Hiring!
nkang@databricks.com

More Related Content

What's hot

Self-Service Apache Spark Structured Streaming Applications and Analytics
Self-Service Apache Spark Structured Streaming Applications and AnalyticsSelf-Service Apache Spark Structured Streaming Applications and Analytics
Self-Service Apache Spark Structured Streaming Applications and AnalyticsDatabricks
 
CERN’s Next Generation Data Analysis Platform with Apache Spark with Enric Te...
CERN’s Next Generation Data Analysis Platform with Apache Spark with Enric Te...CERN’s Next Generation Data Analysis Platform with Apache Spark with Enric Te...
CERN’s Next Generation Data Analysis Platform with Apache Spark with Enric Te...Databricks
 
Maximize the Business Value of Machine Learning and Data Science with Kafka (...
Maximize the Business Value of Machine Learning and Data Science with Kafka (...Maximize the Business Value of Machine Learning and Data Science with Kafka (...
Maximize the Business Value of Machine Learning and Data Science with Kafka (...confluent
 
Spark Summit EU talk by John Musser
Spark Summit EU talk by John MusserSpark Summit EU talk by John Musser
Spark Summit EU talk by John MusserSpark Summit
 
Analytical DBMS to Apache Spark Auto Migration Framework with Edward Zhang an...
Analytical DBMS to Apache Spark Auto Migration Framework with Edward Zhang an...Analytical DBMS to Apache Spark Auto Migration Framework with Edward Zhang an...
Analytical DBMS to Apache Spark Auto Migration Framework with Edward Zhang an...Databricks
 
Kafka Lag Monitoring For Human Beings (Elad Leev, AppsFlyer) Kafka Summit 2020
Kafka Lag Monitoring For Human Beings (Elad Leev, AppsFlyer) Kafka Summit 2020Kafka Lag Monitoring For Human Beings (Elad Leev, AppsFlyer) Kafka Summit 2020
Kafka Lag Monitoring For Human Beings (Elad Leev, AppsFlyer) Kafka Summit 2020HostedbyConfluent
 
Building a Streaming Microservice Architecture: with Apache Spark Structured ...
Building a Streaming Microservice Architecture: with Apache Spark Structured ...Building a Streaming Microservice Architecture: with Apache Spark Structured ...
Building a Streaming Microservice Architecture: with Apache Spark Structured ...Databricks
 
OAP: Optimized Analytics Package for Spark Platform with Daoyuan Wang and Yua...
OAP: Optimized Analytics Package for Spark Platform with Daoyuan Wang and Yua...OAP: Optimized Analytics Package for Spark Platform with Daoyuan Wang and Yua...
OAP: Optimized Analytics Package for Spark Platform with Daoyuan Wang and Yua...Databricks
 
Spark Summit EU talk by Kaarthik Sivashanmugam
Spark Summit EU talk by Kaarthik SivashanmugamSpark Summit EU talk by Kaarthik Sivashanmugam
Spark Summit EU talk by Kaarthik SivashanmugamSpark Summit
 
Efficient State Management With Spark 2.0 And Scale-Out Databases
Efficient State Management With Spark 2.0 And Scale-Out DatabasesEfficient State Management With Spark 2.0 And Scale-Out Databases
Efficient State Management With Spark 2.0 And Scale-Out DatabasesJen Aman
 
Monitoring and Troubleshooting a Real Time Pipeline
Monitoring and Troubleshooting a Real Time PipelineMonitoring and Troubleshooting a Real Time Pipeline
Monitoring and Troubleshooting a Real Time PipelineApache Apex
 
Spark Operator—Deploy, Manage and Monitor Spark clusters on Kubernetes
 Spark Operator—Deploy, Manage and Monitor Spark clusters on Kubernetes Spark Operator—Deploy, Manage and Monitor Spark clusters on Kubernetes
Spark Operator—Deploy, Manage and Monitor Spark clusters on KubernetesDatabricks
 
Building Continuous Application with Structured Streaming and Real-Time Data ...
Building Continuous Application with Structured Streaming and Real-Time Data ...Building Continuous Application with Structured Streaming and Real-Time Data ...
Building Continuous Application with Structured Streaming and Real-Time Data ...Databricks
 
Volta: Logging, Metrics, and Monitoring as a Service
Volta: Logging, Metrics, and Monitoring as a ServiceVolta: Logging, Metrics, and Monitoring as a Service
Volta: Logging, Metrics, and Monitoring as a ServiceLN Renganarayana
 
Kafka: Journey from Just Another Software to Being a Critical Part of PayPal ...
Kafka: Journey from Just Another Software to Being a Critical Part of PayPal ...Kafka: Journey from Just Another Software to Being a Critical Part of PayPal ...
Kafka: Journey from Just Another Software to Being a Critical Part of PayPal ...confluent
 
Case Study: Elasticsearch Ingest Using StreamSets at Cisco Intercloud
Case Study: Elasticsearch Ingest Using StreamSets at Cisco IntercloudCase Study: Elasticsearch Ingest Using StreamSets at Cisco Intercloud
Case Study: Elasticsearch Ingest Using StreamSets at Cisco IntercloudRick Bilodeau
 
SingleStore & Kafka: Better Together to Power Modern Real-Time Data Architect...
SingleStore & Kafka: Better Together to Power Modern Real-Time Data Architect...SingleStore & Kafka: Better Together to Power Modern Real-Time Data Architect...
SingleStore & Kafka: Better Together to Power Modern Real-Time Data Architect...HostedbyConfluent
 
Dr. Elephant for Monitoring and Tuning Apache Spark Jobs on Hadoop with Carl ...
Dr. Elephant for Monitoring and Tuning Apache Spark Jobs on Hadoop with Carl ...Dr. Elephant for Monitoring and Tuning Apache Spark Jobs on Hadoop with Carl ...
Dr. Elephant for Monitoring and Tuning Apache Spark Jobs on Hadoop with Carl ...Databricks
 
Data Policies for the Kafka-API with WebAssembly | Alexander Gallego, Vectorized
Data Policies for the Kafka-API with WebAssembly | Alexander Gallego, VectorizedData Policies for the Kafka-API with WebAssembly | Alexander Gallego, Vectorized
Data Policies for the Kafka-API with WebAssembly | Alexander Gallego, VectorizedHostedbyConfluent
 

What's hot (20)

Self-Service Apache Spark Structured Streaming Applications and Analytics
Self-Service Apache Spark Structured Streaming Applications and AnalyticsSelf-Service Apache Spark Structured Streaming Applications and Analytics
Self-Service Apache Spark Structured Streaming Applications and Analytics
 
CERN’s Next Generation Data Analysis Platform with Apache Spark with Enric Te...
CERN’s Next Generation Data Analysis Platform with Apache Spark with Enric Te...CERN’s Next Generation Data Analysis Platform with Apache Spark with Enric Te...
CERN’s Next Generation Data Analysis Platform with Apache Spark with Enric Te...
 
Maximize the Business Value of Machine Learning and Data Science with Kafka (...
Maximize the Business Value of Machine Learning and Data Science with Kafka (...Maximize the Business Value of Machine Learning and Data Science with Kafka (...
Maximize the Business Value of Machine Learning and Data Science with Kafka (...
 
Spark Summit EU talk by John Musser
Spark Summit EU talk by John MusserSpark Summit EU talk by John Musser
Spark Summit EU talk by John Musser
 
Data Pipeline at Tapad
Data Pipeline at TapadData Pipeline at Tapad
Data Pipeline at Tapad
 
Analytical DBMS to Apache Spark Auto Migration Framework with Edward Zhang an...
Analytical DBMS to Apache Spark Auto Migration Framework with Edward Zhang an...Analytical DBMS to Apache Spark Auto Migration Framework with Edward Zhang an...
Analytical DBMS to Apache Spark Auto Migration Framework with Edward Zhang an...
 
Kafka Lag Monitoring For Human Beings (Elad Leev, AppsFlyer) Kafka Summit 2020
Kafka Lag Monitoring For Human Beings (Elad Leev, AppsFlyer) Kafka Summit 2020Kafka Lag Monitoring For Human Beings (Elad Leev, AppsFlyer) Kafka Summit 2020
Kafka Lag Monitoring For Human Beings (Elad Leev, AppsFlyer) Kafka Summit 2020
 
Building a Streaming Microservice Architecture: with Apache Spark Structured ...
Building a Streaming Microservice Architecture: with Apache Spark Structured ...Building a Streaming Microservice Architecture: with Apache Spark Structured ...
Building a Streaming Microservice Architecture: with Apache Spark Structured ...
 
OAP: Optimized Analytics Package for Spark Platform with Daoyuan Wang and Yua...
OAP: Optimized Analytics Package for Spark Platform with Daoyuan Wang and Yua...OAP: Optimized Analytics Package for Spark Platform with Daoyuan Wang and Yua...
OAP: Optimized Analytics Package for Spark Platform with Daoyuan Wang and Yua...
 
Spark Summit EU talk by Kaarthik Sivashanmugam
Spark Summit EU talk by Kaarthik SivashanmugamSpark Summit EU talk by Kaarthik Sivashanmugam
Spark Summit EU talk by Kaarthik Sivashanmugam
 
Efficient State Management With Spark 2.0 And Scale-Out Databases
Efficient State Management With Spark 2.0 And Scale-Out DatabasesEfficient State Management With Spark 2.0 And Scale-Out Databases
Efficient State Management With Spark 2.0 And Scale-Out Databases
 
Monitoring and Troubleshooting a Real Time Pipeline
Monitoring and Troubleshooting a Real Time PipelineMonitoring and Troubleshooting a Real Time Pipeline
Monitoring and Troubleshooting a Real Time Pipeline
 
Spark Operator—Deploy, Manage and Monitor Spark clusters on Kubernetes
 Spark Operator—Deploy, Manage and Monitor Spark clusters on Kubernetes Spark Operator—Deploy, Manage and Monitor Spark clusters on Kubernetes
Spark Operator—Deploy, Manage and Monitor Spark clusters on Kubernetes
 
Building Continuous Application with Structured Streaming and Real-Time Data ...
Building Continuous Application with Structured Streaming and Real-Time Data ...Building Continuous Application with Structured Streaming and Real-Time Data ...
Building Continuous Application with Structured Streaming and Real-Time Data ...
 
Volta: Logging, Metrics, and Monitoring as a Service
Volta: Logging, Metrics, and Monitoring as a ServiceVolta: Logging, Metrics, and Monitoring as a Service
Volta: Logging, Metrics, and Monitoring as a Service
 
Kafka: Journey from Just Another Software to Being a Critical Part of PayPal ...
Kafka: Journey from Just Another Software to Being a Critical Part of PayPal ...Kafka: Journey from Just Another Software to Being a Critical Part of PayPal ...
Kafka: Journey from Just Another Software to Being a Critical Part of PayPal ...
 
Case Study: Elasticsearch Ingest Using StreamSets at Cisco Intercloud
Case Study: Elasticsearch Ingest Using StreamSets at Cisco IntercloudCase Study: Elasticsearch Ingest Using StreamSets at Cisco Intercloud
Case Study: Elasticsearch Ingest Using StreamSets at Cisco Intercloud
 
SingleStore & Kafka: Better Together to Power Modern Real-Time Data Architect...
SingleStore & Kafka: Better Together to Power Modern Real-Time Data Architect...SingleStore & Kafka: Better Together to Power Modern Real-Time Data Architect...
SingleStore & Kafka: Better Together to Power Modern Real-Time Data Architect...
 
Dr. Elephant for Monitoring and Tuning Apache Spark Jobs on Hadoop with Carl ...
Dr. Elephant for Monitoring and Tuning Apache Spark Jobs on Hadoop with Carl ...Dr. Elephant for Monitoring and Tuning Apache Spark Jobs on Hadoop with Carl ...
Dr. Elephant for Monitoring and Tuning Apache Spark Jobs on Hadoop with Carl ...
 
Data Policies for the Kafka-API with WebAssembly | Alexander Gallego, Vectorized
Data Policies for the Kafka-API with WebAssembly | Alexander Gallego, VectorizedData Policies for the Kafka-API with WebAssembly | Alexander Gallego, Vectorized
Data Policies for the Kafka-API with WebAssembly | Alexander Gallego, Vectorized
 

Viewers also liked

Improving performance of decision support queries in columnar cloud database ...
Improving performance of decision support queries in columnar cloud database ...Improving performance of decision support queries in columnar cloud database ...
Improving performance of decision support queries in columnar cloud database ...Serkan Özal
 
Combining Machine Learning Frameworks with Apache Spark
Combining Machine Learning Frameworks with Apache SparkCombining Machine Learning Frameworks with Apache Spark
Combining Machine Learning Frameworks with Apache SparkDatabricks
 
A Tale of Three Tools: Kubernetes, Jsonnet, and Bazel
A Tale of Three Tools: Kubernetes, Jsonnet, and BazelA Tale of Three Tools: Kubernetes, Jsonnet, and Bazel
A Tale of Three Tools: Kubernetes, Jsonnet, and BazelDatabricks
 
AWS re:Invent 2016: Deep Dive: Amazon EMR Best Practices & Design Patterns (B...
AWS re:Invent 2016: Deep Dive: Amazon EMR Best Practices & Design Patterns (B...AWS re:Invent 2016: Deep Dive: Amazon EMR Best Practices & Design Patterns (B...
AWS re:Invent 2016: Deep Dive: Amazon EMR Best Practices & Design Patterns (B...Amazon Web Services
 
Laying down the smack on your data pipelines
Laying down the smack on your data pipelinesLaying down the smack on your data pipelines
Laying down the smack on your data pipelinesPatrick McFadin
 
Lessons from the Field: Applying Best Practices to Your Apache Spark Applicat...
Lessons from the Field: Applying Best Practices to Your Apache Spark Applicat...Lessons from the Field: Applying Best Practices to Your Apache Spark Applicat...
Lessons from the Field: Applying Best Practices to Your Apache Spark Applicat...Databricks
 
A Developer's View Into Spark's Memory Model with Wenchen Fan
A Developer's View Into Spark's Memory Model with Wenchen FanA Developer's View Into Spark's Memory Model with Wenchen Fan
A Developer's View Into Spark's Memory Model with Wenchen FanDatabricks
 

Viewers also liked (7)

Improving performance of decision support queries in columnar cloud database ...
Improving performance of decision support queries in columnar cloud database ...Improving performance of decision support queries in columnar cloud database ...
Improving performance of decision support queries in columnar cloud database ...
 
Combining Machine Learning Frameworks with Apache Spark
Combining Machine Learning Frameworks with Apache SparkCombining Machine Learning Frameworks with Apache Spark
Combining Machine Learning Frameworks with Apache Spark
 
A Tale of Three Tools: Kubernetes, Jsonnet, and Bazel
A Tale of Three Tools: Kubernetes, Jsonnet, and BazelA Tale of Three Tools: Kubernetes, Jsonnet, and Bazel
A Tale of Three Tools: Kubernetes, Jsonnet, and Bazel
 
AWS re:Invent 2016: Deep Dive: Amazon EMR Best Practices & Design Patterns (B...
AWS re:Invent 2016: Deep Dive: Amazon EMR Best Practices & Design Patterns (B...AWS re:Invent 2016: Deep Dive: Amazon EMR Best Practices & Design Patterns (B...
AWS re:Invent 2016: Deep Dive: Amazon EMR Best Practices & Design Patterns (B...
 
Laying down the smack on your data pipelines
Laying down the smack on your data pipelinesLaying down the smack on your data pipelines
Laying down the smack on your data pipelines
 
Lessons from the Field: Applying Best Practices to Your Apache Spark Applicat...
Lessons from the Field: Applying Best Practices to Your Apache Spark Applicat...Lessons from the Field: Applying Best Practices to Your Apache Spark Applicat...
Lessons from the Field: Applying Best Practices to Your Apache Spark Applicat...
 
A Developer's View Into Spark's Memory Model with Wenchen Fan
A Developer's View Into Spark's Memory Model with Wenchen FanA Developer's View Into Spark's Memory Model with Wenchen Fan
A Developer's View Into Spark's Memory Model with Wenchen Fan
 

Similar to Monitoring Large-Scale Apache Spark Clusters at Databricks

Dances with bits - industrial data analytics made easy!
Dances with bits - industrial data analytics made easy!Dances with bits - industrial data analytics made easy!
Dances with bits - industrial data analytics made easy!Julian Feinauer
 
Resilient Predictive Data Pipelines (QCon London 2016)
Resilient Predictive Data Pipelines (QCon London 2016)Resilient Predictive Data Pipelines (QCon London 2016)
Resilient Predictive Data Pipelines (QCon London 2016)Sid Anand
 
Splunk App for Stream
Splunk App for StreamSplunk App for Stream
Splunk App for StreamSplunk
 
The Art of Container Monitoring
The Art of Container MonitoringThe Art of Container Monitoring
The Art of Container MonitoringDerek Chen
 
Building an Observability Platform in 389 Difficult Steps
Building an Observability Platform in 389 Difficult StepsBuilding an Observability Platform in 389 Difficult Steps
Building an Observability Platform in 389 Difficult StepsDigitalOcean
 
AWS re:Invent 2016: How Fulfillment by Amazon (FBA) and Scopely Improved Resu...
AWS re:Invent 2016: How Fulfillment by Amazon (FBA) and Scopely Improved Resu...AWS re:Invent 2016: How Fulfillment by Amazon (FBA) and Scopely Improved Resu...
AWS re:Invent 2016: How Fulfillment by Amazon (FBA) and Scopely Improved Resu...Amazon Web Services
 
Benchmark Showdown: Which Relational Database is the Fastest on AWS?
Benchmark Showdown: Which Relational Database is the Fastest on AWS?Benchmark Showdown: Which Relational Database is the Fastest on AWS?
Benchmark Showdown: Which Relational Database is the Fastest on AWS?Clustrix
 
Apache Pulsar Overview
Apache Pulsar OverviewApache Pulsar Overview
Apache Pulsar OverviewStreamlio
 
Service quality monitoring system architecture
Service quality monitoring system architectureService quality monitoring system architecture
Service quality monitoring system architectureMatsuo Sawahashi
 
Building Streaming And Fast Data Applications With Spark, Mesos, Akka, Cassan...
Building Streaming And Fast Data Applications With Spark, Mesos, Akka, Cassan...Building Streaming And Fast Data Applications With Spark, Mesos, Akka, Cassan...
Building Streaming And Fast Data Applications With Spark, Mesos, Akka, Cassan...Lightbend
 
Monitoring Half a Million ML Models, IoT Streaming Data, and Automated Qualit...
Monitoring Half a Million ML Models, IoT Streaming Data, and Automated Qualit...Monitoring Half a Million ML Models, IoT Streaming Data, and Automated Qualit...
Monitoring Half a Million ML Models, IoT Streaming Data, and Automated Qualit...Databricks
 
Flopsar tesacom-technical-introduction v1a-eng
Flopsar tesacom-technical-introduction v1a-engFlopsar tesacom-technical-introduction v1a-eng
Flopsar tesacom-technical-introduction v1a-engMarcelo Marinangeli
 
Monitoring Containerized Micro-Services In Azure
Monitoring Containerized Micro-Services In AzureMonitoring Containerized Micro-Services In Azure
Monitoring Containerized Micro-Services In AzureAlex Bulankou
 
Kaseya Connect 2012 - THE ABC'S OF MONITORING
Kaseya Connect 2012 - THE ABC'S OF MONITORINGKaseya Connect 2012 - THE ABC'S OF MONITORING
Kaseya Connect 2012 - THE ABC'S OF MONITORINGKaseya
 
Monitoring microservices platform
Monitoring microservices platformMonitoring microservices platform
Monitoring microservices platformBoyan Dimitrov
 
Why & how to optimize sql server for performance from design to query
Why & how to optimize sql server for performance from design to queryWhy & how to optimize sql server for performance from design to query
Why & how to optimize sql server for performance from design to queryAntonios Chatzipavlis
 
John adams talk cloudy
John adams   talk cloudyJohn adams   talk cloudy
John adams talk cloudyJohn Adams
 
Agile Gurugram 2023 | Observability for Modern Applications. How does it help...
Agile Gurugram 2023 | Observability for Modern Applications. How does it help...Agile Gurugram 2023 | Observability for Modern Applications. How does it help...
Agile Gurugram 2023 | Observability for Modern Applications. How does it help...AgileNetwork
 
Monitoring Big Data Systems Done "The Simple Way" - Demi Ben-Ari - Codemotion...
Monitoring Big Data Systems Done "The Simple Way" - Demi Ben-Ari - Codemotion...Monitoring Big Data Systems Done "The Simple Way" - Demi Ben-Ari - Codemotion...
Monitoring Big Data Systems Done "The Simple Way" - Demi Ben-Ari - Codemotion...Codemotion
 
Monitoring Big Data Systems "Done the simple way" - Demi Ben-Ari - Codemotion...
Monitoring Big Data Systems "Done the simple way" - Demi Ben-Ari - Codemotion...Monitoring Big Data Systems "Done the simple way" - Demi Ben-Ari - Codemotion...
Monitoring Big Data Systems "Done the simple way" - Demi Ben-Ari - Codemotion...Demi Ben-Ari
 

Similar to Monitoring Large-Scale Apache Spark Clusters at Databricks (20)

Dances with bits - industrial data analytics made easy!
Dances with bits - industrial data analytics made easy!Dances with bits - industrial data analytics made easy!
Dances with bits - industrial data analytics made easy!
 
Resilient Predictive Data Pipelines (QCon London 2016)
Resilient Predictive Data Pipelines (QCon London 2016)Resilient Predictive Data Pipelines (QCon London 2016)
Resilient Predictive Data Pipelines (QCon London 2016)
 
Splunk App for Stream
Splunk App for StreamSplunk App for Stream
Splunk App for Stream
 
The Art of Container Monitoring
The Art of Container MonitoringThe Art of Container Monitoring
The Art of Container Monitoring
 
Building an Observability Platform in 389 Difficult Steps
Building an Observability Platform in 389 Difficult StepsBuilding an Observability Platform in 389 Difficult Steps
Building an Observability Platform in 389 Difficult Steps
 
AWS re:Invent 2016: How Fulfillment by Amazon (FBA) and Scopely Improved Resu...
AWS re:Invent 2016: How Fulfillment by Amazon (FBA) and Scopely Improved Resu...AWS re:Invent 2016: How Fulfillment by Amazon (FBA) and Scopely Improved Resu...
AWS re:Invent 2016: How Fulfillment by Amazon (FBA) and Scopely Improved Resu...
 
Benchmark Showdown: Which Relational Database is the Fastest on AWS?
Benchmark Showdown: Which Relational Database is the Fastest on AWS?Benchmark Showdown: Which Relational Database is the Fastest on AWS?
Benchmark Showdown: Which Relational Database is the Fastest on AWS?
 
Apache Pulsar Overview
Apache Pulsar OverviewApache Pulsar Overview
Apache Pulsar Overview
 
Service quality monitoring system architecture
Service quality monitoring system architectureService quality monitoring system architecture
Service quality monitoring system architecture
 
Building Streaming And Fast Data Applications With Spark, Mesos, Akka, Cassan...
Building Streaming And Fast Data Applications With Spark, Mesos, Akka, Cassan...Building Streaming And Fast Data Applications With Spark, Mesos, Akka, Cassan...
Building Streaming And Fast Data Applications With Spark, Mesos, Akka, Cassan...
 
Monitoring Half a Million ML Models, IoT Streaming Data, and Automated Qualit...
Monitoring Half a Million ML Models, IoT Streaming Data, and Automated Qualit...Monitoring Half a Million ML Models, IoT Streaming Data, and Automated Qualit...
Monitoring Half a Million ML Models, IoT Streaming Data, and Automated Qualit...
 
Flopsar tesacom-technical-introduction v1a-eng
Flopsar tesacom-technical-introduction v1a-engFlopsar tesacom-technical-introduction v1a-eng
Flopsar tesacom-technical-introduction v1a-eng
 
Monitoring Containerized Micro-Services In Azure
Monitoring Containerized Micro-Services In AzureMonitoring Containerized Micro-Services In Azure
Monitoring Containerized Micro-Services In Azure
 
Kaseya Connect 2012 - THE ABC'S OF MONITORING
Kaseya Connect 2012 - THE ABC'S OF MONITORINGKaseya Connect 2012 - THE ABC'S OF MONITORING
Kaseya Connect 2012 - THE ABC'S OF MONITORING
 
Monitoring microservices platform
Monitoring microservices platformMonitoring microservices platform
Monitoring microservices platform
 
Why & how to optimize sql server for performance from design to query
Why & how to optimize sql server for performance from design to queryWhy & how to optimize sql server for performance from design to query
Why & how to optimize sql server for performance from design to query
 
John adams talk cloudy
John adams   talk cloudyJohn adams   talk cloudy
John adams talk cloudy
 
Agile Gurugram 2023 | Observability for Modern Applications. How does it help...
Agile Gurugram 2023 | Observability for Modern Applications. How does it help...Agile Gurugram 2023 | Observability for Modern Applications. How does it help...
Agile Gurugram 2023 | Observability for Modern Applications. How does it help...
 
Monitoring Big Data Systems Done "The Simple Way" - Demi Ben-Ari - Codemotion...
Monitoring Big Data Systems Done "The Simple Way" - Demi Ben-Ari - Codemotion...Monitoring Big Data Systems Done "The Simple Way" - Demi Ben-Ari - Codemotion...
Monitoring Big Data Systems Done "The Simple Way" - Demi Ben-Ari - Codemotion...
 
Monitoring Big Data Systems "Done the simple way" - Demi Ben-Ari - Codemotion...
Monitoring Big Data Systems "Done the simple way" - Demi Ben-Ari - Codemotion...Monitoring Big Data Systems "Done the simple way" - Demi Ben-Ari - Codemotion...
Monitoring Big Data Systems "Done the simple way" - Demi Ben-Ari - Codemotion...
 

More from Anyscale

ACM Sunnyvale Meetup.pdf
ACM Sunnyvale Meetup.pdfACM Sunnyvale Meetup.pdf
ACM Sunnyvale Meetup.pdfAnyscale
 
What's Next for MLflow in 2019
What's Next for MLflow in 2019What's Next for MLflow in 2019
What's Next for MLflow in 2019Anyscale
 
Putting AI to Work on Apache Spark
Putting AI to Work on Apache SparkPutting AI to Work on Apache Spark
Putting AI to Work on Apache SparkAnyscale
 
Project Hydrogen, HorovodRunner, and Pandas UDF: Distributed Deep Learning Tr...
Project Hydrogen, HorovodRunner, and Pandas UDF: Distributed Deep Learning Tr...Project Hydrogen, HorovodRunner, and Pandas UDF: Distributed Deep Learning Tr...
Project Hydrogen, HorovodRunner, and Pandas UDF: Distributed Deep Learning Tr...Anyscale
 
Jump Start on Apache Spark 2.2 with Databricks
Jump Start on Apache Spark 2.2 with DatabricksJump Start on Apache Spark 2.2 with Databricks
Jump Start on Apache Spark 2.2 with DatabricksAnyscale
 
Easy, scalable, fault tolerant stream processing with structured streaming - ...
Easy, scalable, fault tolerant stream processing with structured streaming - ...Easy, scalable, fault tolerant stream processing with structured streaming - ...
Easy, scalable, fault tolerant stream processing with structured streaming - ...Anyscale
 
Apache ® Spark™ MLlib 2.x: How to Productionize your Machine Learning Models
Apache ® Spark™ MLlib 2.x: How to Productionize your Machine Learning ModelsApache ® Spark™ MLlib 2.x: How to Productionize your Machine Learning Models
Apache ® Spark™ MLlib 2.x: How to Productionize your Machine Learning ModelsAnyscale
 
A Deep Dive into Structured Streaming in Apache Spark
A Deep Dive into Structured Streaming in Apache Spark A Deep Dive into Structured Streaming in Apache Spark
A Deep Dive into Structured Streaming in Apache Spark Anyscale
 
Jump Start with Apache Spark 2.0 on Databricks
Jump Start with Apache Spark 2.0 on DatabricksJump Start with Apache Spark 2.0 on Databricks
Jump Start with Apache Spark 2.0 on DatabricksAnyscale
 
Continuous Application with Structured Streaming 2.0
Continuous Application with Structured Streaming 2.0Continuous Application with Structured Streaming 2.0
Continuous Application with Structured Streaming 2.0Anyscale
 

More from Anyscale (10)

ACM Sunnyvale Meetup.pdf
ACM Sunnyvale Meetup.pdfACM Sunnyvale Meetup.pdf
ACM Sunnyvale Meetup.pdf
 
What's Next for MLflow in 2019
What's Next for MLflow in 2019What's Next for MLflow in 2019
What's Next for MLflow in 2019
 
Putting AI to Work on Apache Spark
Putting AI to Work on Apache SparkPutting AI to Work on Apache Spark
Putting AI to Work on Apache Spark
 
Project Hydrogen, HorovodRunner, and Pandas UDF: Distributed Deep Learning Tr...
Project Hydrogen, HorovodRunner, and Pandas UDF: Distributed Deep Learning Tr...Project Hydrogen, HorovodRunner, and Pandas UDF: Distributed Deep Learning Tr...
Project Hydrogen, HorovodRunner, and Pandas UDF: Distributed Deep Learning Tr...
 
Jump Start on Apache Spark 2.2 with Databricks
Jump Start on Apache Spark 2.2 with DatabricksJump Start on Apache Spark 2.2 with Databricks
Jump Start on Apache Spark 2.2 with Databricks
 
Easy, scalable, fault tolerant stream processing with structured streaming - ...
Easy, scalable, fault tolerant stream processing with structured streaming - ...Easy, scalable, fault tolerant stream processing with structured streaming - ...
Easy, scalable, fault tolerant stream processing with structured streaming - ...
 
Apache ® Spark™ MLlib 2.x: How to Productionize your Machine Learning Models
Apache ® Spark™ MLlib 2.x: How to Productionize your Machine Learning ModelsApache ® Spark™ MLlib 2.x: How to Productionize your Machine Learning Models
Apache ® Spark™ MLlib 2.x: How to Productionize your Machine Learning Models
 
A Deep Dive into Structured Streaming in Apache Spark
A Deep Dive into Structured Streaming in Apache Spark A Deep Dive into Structured Streaming in Apache Spark
A Deep Dive into Structured Streaming in Apache Spark
 
Jump Start with Apache Spark 2.0 on Databricks
Jump Start with Apache Spark 2.0 on DatabricksJump Start with Apache Spark 2.0 on Databricks
Jump Start with Apache Spark 2.0 on Databricks
 
Continuous Application with Structured Streaming 2.0
Continuous Application with Structured Streaming 2.0Continuous Application with Structured Streaming 2.0
Continuous Application with Structured Streaming 2.0
 

Recently uploaded

Virtual memory management in Operating System
Virtual memory management in Operating SystemVirtual memory management in Operating System
Virtual memory management in Operating SystemRashmi Bhat
 
Cost estimation approach: FP to COCOMO scenario based question
Cost estimation approach: FP to COCOMO scenario based questionCost estimation approach: FP to COCOMO scenario based question
Cost estimation approach: FP to COCOMO scenario based questionSneha Padhiar
 
THE SENDAI FRAMEWORK FOR DISASTER RISK REDUCTION
THE SENDAI FRAMEWORK FOR DISASTER RISK REDUCTIONTHE SENDAI FRAMEWORK FOR DISASTER RISK REDUCTION
THE SENDAI FRAMEWORK FOR DISASTER RISK REDUCTIONjhunlian
 
Research Methodology for Engineering pdf
Research Methodology for Engineering pdfResearch Methodology for Engineering pdf
Research Methodology for Engineering pdfCaalaaAbdulkerim
 
CS 3251 Programming in c all unit notes pdf
CS 3251 Programming in c all unit notes pdfCS 3251 Programming in c all unit notes pdf
CS 3251 Programming in c all unit notes pdfBalamuruganV28
 
2022 AWS DNA Hackathon 장애 대응 솔루션 jarvis.
2022 AWS DNA Hackathon 장애 대응 솔루션 jarvis.2022 AWS DNA Hackathon 장애 대응 솔루션 jarvis.
2022 AWS DNA Hackathon 장애 대응 솔루션 jarvis.elesangwon
 
Robotics-Asimov's Laws, Mechanical Subsystems, Robot Kinematics, Robot Dynami...
Robotics-Asimov's Laws, Mechanical Subsystems, Robot Kinematics, Robot Dynami...Robotics-Asimov's Laws, Mechanical Subsystems, Robot Kinematics, Robot Dynami...
Robotics-Asimov's Laws, Mechanical Subsystems, Robot Kinematics, Robot Dynami...Sumanth A
 
FUNCTIONAL AND NON FUNCTIONAL REQUIREMENT
FUNCTIONAL AND NON FUNCTIONAL REQUIREMENTFUNCTIONAL AND NON FUNCTIONAL REQUIREMENT
FUNCTIONAL AND NON FUNCTIONAL REQUIREMENTSneha Padhiar
 
Katarzyna Lipka-Sidor - BIM School Course
Katarzyna Lipka-Sidor - BIM School CourseKatarzyna Lipka-Sidor - BIM School Course
Katarzyna Lipka-Sidor - BIM School Coursebim.edu.pl
 
Gravity concentration_MI20612MI_________
Gravity concentration_MI20612MI_________Gravity concentration_MI20612MI_________
Gravity concentration_MI20612MI_________Romil Mishra
 
Engineering Drawing section of solid
Engineering Drawing     section of solidEngineering Drawing     section of solid
Engineering Drawing section of solidnamansinghjarodiya
 
Novel 3D-Printed Soft Linear and Bending Actuators
Novel 3D-Printed Soft Linear and Bending ActuatorsNovel 3D-Printed Soft Linear and Bending Actuators
Novel 3D-Printed Soft Linear and Bending ActuatorsResearcher Researcher
 
DEVICE DRIVERS AND INTERRUPTS SERVICE MECHANISM.pdf
DEVICE DRIVERS AND INTERRUPTS  SERVICE MECHANISM.pdfDEVICE DRIVERS AND INTERRUPTS  SERVICE MECHANISM.pdf
DEVICE DRIVERS AND INTERRUPTS SERVICE MECHANISM.pdfAkritiPradhan2
 
Ch10-Global Supply Chain - Cadena de Suministro.pdf
Ch10-Global Supply Chain - Cadena de Suministro.pdfCh10-Global Supply Chain - Cadena de Suministro.pdf
Ch10-Global Supply Chain - Cadena de Suministro.pdfChristianCDAM
 
Comprehensive energy systems.pdf Comprehensive energy systems.pdf
Comprehensive energy systems.pdf Comprehensive energy systems.pdfComprehensive energy systems.pdf Comprehensive energy systems.pdf
Comprehensive energy systems.pdf Comprehensive energy systems.pdfalene1
 
Module-1-(Building Acoustics) Noise Control (Unit-3). pdf
Module-1-(Building Acoustics) Noise Control (Unit-3). pdfModule-1-(Building Acoustics) Noise Control (Unit-3). pdf
Module-1-(Building Acoustics) Noise Control (Unit-3). pdfManish Kumar
 
11. Properties of Liquid Fuels in Energy Engineering.pdf
11. Properties of Liquid Fuels in Energy Engineering.pdf11. Properties of Liquid Fuels in Energy Engineering.pdf
11. Properties of Liquid Fuels in Energy Engineering.pdfHafizMudaserAhmad
 
OOP concepts -in-Python programming language
OOP concepts -in-Python programming languageOOP concepts -in-Python programming language
OOP concepts -in-Python programming languageSmritiSharma901052
 
Stork Webinar | APM Transformational planning, Tool Selection & Performance T...
Stork Webinar | APM Transformational planning, Tool Selection & Performance T...Stork Webinar | APM Transformational planning, Tool Selection & Performance T...
Stork Webinar | APM Transformational planning, Tool Selection & Performance T...Stork
 
List of Accredited Concrete Batching Plant.pdf
List of Accredited Concrete Batching Plant.pdfList of Accredited Concrete Batching Plant.pdf
List of Accredited Concrete Batching Plant.pdfisabel213075
 

Recently uploaded (20)

Virtual memory management in Operating System
Virtual memory management in Operating SystemVirtual memory management in Operating System
Virtual memory management in Operating System
 
Cost estimation approach: FP to COCOMO scenario based question
Cost estimation approach: FP to COCOMO scenario based questionCost estimation approach: FP to COCOMO scenario based question
Cost estimation approach: FP to COCOMO scenario based question
 
THE SENDAI FRAMEWORK FOR DISASTER RISK REDUCTION
THE SENDAI FRAMEWORK FOR DISASTER RISK REDUCTIONTHE SENDAI FRAMEWORK FOR DISASTER RISK REDUCTION
THE SENDAI FRAMEWORK FOR DISASTER RISK REDUCTION
 
Research Methodology for Engineering pdf
Research Methodology for Engineering pdfResearch Methodology for Engineering pdf
Research Methodology for Engineering pdf
 
CS 3251 Programming in c all unit notes pdf
CS 3251 Programming in c all unit notes pdfCS 3251 Programming in c all unit notes pdf
CS 3251 Programming in c all unit notes pdf
 
2022 AWS DNA Hackathon 장애 대응 솔루션 jarvis.
2022 AWS DNA Hackathon 장애 대응 솔루션 jarvis.2022 AWS DNA Hackathon 장애 대응 솔루션 jarvis.
2022 AWS DNA Hackathon 장애 대응 솔루션 jarvis.
 
Robotics-Asimov's Laws, Mechanical Subsystems, Robot Kinematics, Robot Dynami...
Robotics-Asimov's Laws, Mechanical Subsystems, Robot Kinematics, Robot Dynami...Robotics-Asimov's Laws, Mechanical Subsystems, Robot Kinematics, Robot Dynami...
Robotics-Asimov's Laws, Mechanical Subsystems, Robot Kinematics, Robot Dynami...
 
FUNCTIONAL AND NON FUNCTIONAL REQUIREMENT
FUNCTIONAL AND NON FUNCTIONAL REQUIREMENTFUNCTIONAL AND NON FUNCTIONAL REQUIREMENT
FUNCTIONAL AND NON FUNCTIONAL REQUIREMENT
 
Katarzyna Lipka-Sidor - BIM School Course
Katarzyna Lipka-Sidor - BIM School CourseKatarzyna Lipka-Sidor - BIM School Course
Katarzyna Lipka-Sidor - BIM School Course
 
Gravity concentration_MI20612MI_________
Gravity concentration_MI20612MI_________Gravity concentration_MI20612MI_________
Gravity concentration_MI20612MI_________
 
Engineering Drawing section of solid
Engineering Drawing     section of solidEngineering Drawing     section of solid
Engineering Drawing section of solid
 
Novel 3D-Printed Soft Linear and Bending Actuators
Novel 3D-Printed Soft Linear and Bending ActuatorsNovel 3D-Printed Soft Linear and Bending Actuators
Novel 3D-Printed Soft Linear and Bending Actuators
 
DEVICE DRIVERS AND INTERRUPTS SERVICE MECHANISM.pdf
DEVICE DRIVERS AND INTERRUPTS  SERVICE MECHANISM.pdfDEVICE DRIVERS AND INTERRUPTS  SERVICE MECHANISM.pdf
DEVICE DRIVERS AND INTERRUPTS SERVICE MECHANISM.pdf
 
Ch10-Global Supply Chain - Cadena de Suministro.pdf
Ch10-Global Supply Chain - Cadena de Suministro.pdfCh10-Global Supply Chain - Cadena de Suministro.pdf
Ch10-Global Supply Chain - Cadena de Suministro.pdf
 
Comprehensive energy systems.pdf Comprehensive energy systems.pdf
Comprehensive energy systems.pdf Comprehensive energy systems.pdfComprehensive energy systems.pdf Comprehensive energy systems.pdf
Comprehensive energy systems.pdf Comprehensive energy systems.pdf
 
Module-1-(Building Acoustics) Noise Control (Unit-3). pdf
Module-1-(Building Acoustics) Noise Control (Unit-3). pdfModule-1-(Building Acoustics) Noise Control (Unit-3). pdf
Module-1-(Building Acoustics) Noise Control (Unit-3). pdf
 
11. Properties of Liquid Fuels in Energy Engineering.pdf
11. Properties of Liquid Fuels in Energy Engineering.pdf11. Properties of Liquid Fuels in Energy Engineering.pdf
11. Properties of Liquid Fuels in Energy Engineering.pdf
 
OOP concepts -in-Python programming language
OOP concepts -in-Python programming languageOOP concepts -in-Python programming language
OOP concepts -in-Python programming language
 
Stork Webinar | APM Transformational planning, Tool Selection & Performance T...
Stork Webinar | APM Transformational planning, Tool Selection & Performance T...Stork Webinar | APM Transformational planning, Tool Selection & Performance T...
Stork Webinar | APM Transformational planning, Tool Selection & Performance T...
 
List of Accredited Concrete Batching Plant.pdf
List of Accredited Concrete Batching Plant.pdfList of Accredited Concrete Batching Plant.pdf
List of Accredited Concrete Batching Plant.pdf
 

Monitoring Large-Scale Apache Spark Clusters at Databricks

  • 1. Monitoring Large-Scale Spark Clusters at Databricks Nanxi Kang Apr.12th
  • 2. 2 Who we are Inventors of • The largest open source project in data processing • Unified platform for ETL, interactive query, machine learning, streaming, ... Leader in Big Data ecosystem
  • 3. 3 What we do We offer a cloud-based Spark platform • Demo shard
  • 4. 4 Databricks platform Databricks Create Configure Terminate Customer 1 Customer 2 ClusterClusterCluster ClusterCluster service service service customer requests
  • 5. 5 Requirements of monitoring As an engineer, I want to • Understand the state of the service • Have customer-oriented metrics (uptime, performance regression, ...) • Be alerted on abnormality • Ease debugging and troubleshooting
  • 6. 6 Requirements of monitoring As an engineer, I want to • Understand the state of the service • Have customer-oriented metrics (uptime, performance regression, ...) • Be alerted on abnormality • Ease debugging and troubleshooting Technically • Ingest metrics with low latency • Aggregate metrics • Scalability
  • 7. 7 Ingest and aggregate metrics Prometheus Grafana Alert Manager + PagerDuty Databricks service service Metrics latency{operation=”runCommand”} = 60 alert if latency{operation=”runCommand”} > 100
  • 8. 8 Ingest and aggregate metrics Prometheus Grafana Alert Manager + PagerDuty Databricks service service Metrics Kinesis
  • 9. 9 Latency Prometheus Grafana Alert Manager + PagerDuty Databricks service service Metrics Kinesis ~ 30 seconds ~ 5 seconds refresh every 1 ~ 5 mins ~ 5 seconds
  • 10. 10 Recap Databricks platform Databricks Create Configure Terminate Customer 1 Customer 2 ClusterClusterCluster ClusterCluster service service service customer requests
  • 11. 11 Unique challenges • Heterogeneous environments • Scalability: large volume of metrics – ~300K/min from Databricks – ~2M/min from Customers Customer 1 Databricks ClusterClusterCluster service service service Prometheus
  • 12. 12 Ingest metrics from customer env Prometheus Kinesis Customer
  • 13. 13 Ingest Metrics from Customer Env Prometheus Grafana Alert Manager + PagerDuty Kinesis Customer Kinesis2Prom
  • 14. 14 What Next? Let’s collect metrics! Services can emit structured data about how they’re doing • This is called whitebox monitoring • Example: “how long does it take to finish a spark command?” • Golden metrics (Latency, Traffic, Error, and Saturation)
  • 15. 15 Whitebox monitoring During a recent AWS S3 outage in us-east-1, our services report failures to create Spark clusters.
  • 16. 16 What’s missing? • Hard to cover the end-to-end (cross-)service behaviors • Whitebox metrics are great for causes, not symptoms
  • 17. 17 What’s missing? • Hard to cover the end-to-end (cross-)service behaviors • Whitebox metrics are great for causes, not symptoms Understand the user’s experience • Use a fake user to record the service’s behavior • This is called Blackbox monitoring • Example: a user records the time from “he hits the command run UI button” to “he sees the results/errors from UI.”
  • 19. 19 Our monitoring pipeline needs to be monitored too! Meta-monitoring Prometheus Grafana Alert Manager + PagerDuty Meta-Prometheus
  • 20. 20 We can shard metrics and launch more Prometheus(es). Scale out monitoring pipeline Prometheus Databricks Prometheus Kinesis2Prom metric filters
  • 21. 21 We can shard metrics and launch more Prometheus(es). However, we cannot join metrics across partitions. Scale out monitoring pipeline
  • 22. 22 We can shard metrics and launch more Prometheus(es). However, we cannot join metrics across partitions. We are starting to offload extremely high-volume metrics to Spark’s structured streaming. Scale out monitoring pipeline
  • 23. 23 Monitoring is useful! - Since rollout, all teams at Databricks started using dashboards and alerts Add metrics as you build features Building scalable and robust monitoring systems is hard Take-aways
  • 24. Thanks & We Are Hiring! nkang@databricks.com