Cloudera Customer Success Story
0 likes1,330 views
This document describes a customer success story involving Cloudera and Xpand IT. It discusses how Xpand IT developed a solution to provide near real-time monitoring and management of Hadoop clusters. The solution involves collecting telemetry data from Hadoop jobs, storing it in Kafka for real-time access, and using Spark to parse the logs and load data into Impala and HBase. This allows for real-time monitoring and control of ETL jobs across multiple Hadoop components in a fault-tolerant manner. The architecture is designed according to lambda architecture principles to handle both real-time and batch data processing.
Cloudera Customer Success Story
- 1. Customer Success Story Cloudera & Xpand IT Nuno Barreto Associate Partner & Big Data Lead [email protected] Proprietary & Confidential www.xpand-it.com
- 2. THE PROBLEM How is process Y progressing? Who are the main cluster users/departments? Which engines does each department use? Do I need to plan on an upgrade? How much is process X costing me? Are there available time slots?
- 3. THE SOLUTION TELEMETRY ETL FLOW CONTROL DATA PREPARATION
- 4. ARCHITECURE CORE AGENT(s) QUEUE REAL-TIME ONLINEDB ANALYTICSREPO ETL start/stop jobs start/stop jobs PDI extensionlogflow control ANALYTICS ANALYTICSDB status check metadata access data access analytics data analytical queries operational data
- 5. THIS INVOLES A NUMBER OF CONCEPTS NEAR REAL-TIME CLOUDERA INTEGRATION LAMBDA ARCHITECTURE STREAMING
- 7. REAL-TIME & STREAMING CORE AGENT(s) QUEUE REAL-TIME ONLINEDB ANALYTICSREPO ETL start/stop jobs start/stop jobs PDI extensionlogflow control ANALYTICS ANALYTICSDB status check metadata access data access analytics data analytical queries operational data
- 8. REMOTE AGENTS FINE GRAINED CONTROL ETL TOOL SPECIFIC REAL TIME LOGGING ASYNC EXECUTION
- 9. PDI EXTENSION POINTS CAPTURE LOG START/END CAPTURE CONNECTION TYPE CAPTURE STEP LINEAGE DETAIL
- 10. GATHERING EXECUTION DATA USE KAFKA AS A LOG SINK FAULT TOLERANT REAL TIME CONSISTENT
- 11. COLLECT LOG DATA IN (AS) REALTIME (AS POSSIBLE) SPARK AS KAFKA COLLECTOR REAL TIME LOG PARSING ETL TOOL ADAPTABLE DATA DUMPS IN IMPALA AND HBASE GENERATES NOTIFICATIONS
- 13. LAMBDA ARCHITECTURE CORE AGENT(s) QUEUE REAL-TIME ONLINEDB ANALYTICSREPO ETL start/stop jobs start/stop jobs PDI extensionlogflow control ANALYTICS ANALYTICSDB status check metadata access data access analytics data analytical queries operational data
- 14. DISCLAIMER What you are about to see is a Work In Progress so, be gentle in case… • the demo doesn’t work • features don’t work as described • connection goes down
- 17. HOW TO MANAGE ALL THESE COMPONENTS LOTS OF MOVING PARTS OPERATIONS LOADS OF CONFIG FILES
- 18. THE ANSWER
- 19. EXTENSIBLE ARCHITECTURE SEAMLESS INTEGRATION MONITORING CONFIGURATION MANAGEMENT DEPENDENCIES MANAGEMENT LOG CHECK SETUP AND ADMIN
- 21. SUMMARY NOT EVERYTING WE DO IS THIS COMPLEX HADOOP STACK CHOICE MATTERS RE-USABLE DESIGN PATTERNS
- 22. QUESTIONS?