How to Manage Your Apache Hadoop Lifecycle.
So you’ve got Apache Hadoop in development. Now what? In this webinar, Cloudera’s VP of Products Charles Zedlewski will explain how to plan for and manage the Apache Hadoop lifecycle inside a Cloudera deployment.
11. First principles – set business goals What are the business outcomes Hadoop is supposed to deliver? New insights Lower business costs Lower IT costs More data under management More revenue through better targeting, conversion ? 6 Copyright 2011 Cloudera Inc. All rights reserved
12. First principles – set operations goals Performance Utilization Cost of operations Availability Quality of service Flexibility / elasticity Security Transparency ? 7 Copyright 2011 Cloudera Inc. All rights reserved
13. System design – stick to the basics Hadoop needs to know where it’s hard drives are Running on a virtualized layer is a bad idea RAID is a bad idea Running on remote storage the worst idea Servers - prioritize flexibility over bells and whistles How easily will you be able to expand your cluster? How easily can you evolve your core / spindle ratio? How many companies support that exotic chip, card, drive, power supply, etc? Network – prioritize quality over bells & whistles 10G on the backplane is usually unnecessary Plan how to adapt your topology as your cluster grows 8 Copyright 2011 Cloudera Inc. All rights reserved
14. Hadoop in production – the tribe We have a chief that looks out for the tribe Make sure there’s enough fire for everyone Survival of the tribe is still the main concern Job code distinct from the rest of Hadoop Copyright 2011 Cloudera Inc. All rights reserved
15. Train your chief! Unix & DBA backgrounds are both valid starting points 10 Copyright 2011 Cloudera Inc. All rights reserved
16. Then empower your chief! Managing Hadoop requires Sensible selection of hardware Visibility into users, jobs, activities, hardware, operating system, services, logs and more Ability to make changes to configurations, services, patch levels and more In many organizations the chief is precluded from some of these decisions / actions by preexisting policy Take an “appliance mentality” to Hadoop decision making 11 Copyright 2011 Cloudera Inc. All rights reserved
17. Discovery – monitoring & alerting You want to anticipate & alert on: Health checks & status of key nodes (Namenode, Master, etc) Completion & performance of jobs & pipelines (for SLA measurement) System performance & availability Log events (only specific ones) 12 Copyright 2011 Cloudera Inc. All rights reserved
18. Diagnosis 7 lenses into Hadoop, used in combination Service metrics System metrics Configurations Change history Log history Activities, jobs & tasks Stack trace / profiling One lens rarely tells the whole story 13 Copyright 2011 Cloudera Inc. All rights reserved
19. Avoid the scripts Script to run a check Script to import a file Script to preempt a job Script to instrument a daemon Script to…. 14 Copyright 2011 Cloudera Inc. All rights reserved
20. The web of scripts – where it ends 15 Copyright 2011 Cloudera Inc. All rights reserved Nothing ever changes or improves Garish, jerry-rigged Time goes into maintaining scripts, not achieving the objectives One and only one person loves it
21. Hadoop as a standard platform Fire is not a big deal any more. Pollution, congestion, etc a concern More specialized roles Patching, updating, upgrading, configuring and tuning are all distinct Copyright 2011 Cloudera Inc. All rights reserved
22. Optimize – plan for multi-tenancy Definition – ability of disparate groups, users, data and workloads to operate concurrently on 1 logical Hadoop system Multi-tenancy helps you get more of what you really want Better performance Better cost of operations New insights Greater availability Multi-tenancy has some additional considerations 17 Copyright 2011 Cloudera Inc. All rights reserved
23. Optimize – policies for permissions Authentication Don’t talk to strangers Should integrate with existing IT infrastructure Authentication (Kerberos) patches now part of CDH3 Authorization Not everyone can access everything Ex. Production data sets are read-only to quants / analysts. Analysts have home or group directories for derived data sets. Mostly enforced via HDFS permissions; directory structure and organization is critical Not as fine grained as column level access in EDW, RDBMS (but this is coming) 18 Copyright 2011 Cloudera Inc. All rights reserved
24. Optimize – plan for resources Tracking & establishing policies for usage cluster resources Files, bytes and quotas thereof Tasks, memory, IO, CPU, network and scheduling thereof By now you’ve almost certainly graduated to a sophisticated scheduler Policies to prevent bad behavior (e.g. auto-kill) Monitor and track resource utilization across all groups Periodically review queue / pool decisions to improve QOS 19 Copyright 2011 Cloudera Inc. All rights reserved
25. Wrapping it up The operational lifecycle for Hadoop is similar to other systems but Hadoop itself is not The basics are not a good place to get creative Think command center, not man cave Multi-tenancy is an attractive opportunity with some additional operational burdens There’s lots more work to do 20 Copyright 2011 Cloudera Inc. All rights reserved