my compilation of the changes and differences of the upcoming 3.0 version of Hadoop. Present during the Meetup of the group https://www.meetup.com/Big-Data-Hadoop-Spark-NRW/

1. 1
What‘s new in
Hadoop 3.0
Heiko Loewe
@loeweh

2. 2
Tributes
• Zhe Zhang, Erasure Coding
• Akira Ajisaka, Script Rewrite
• Junping Du, Yarn Timeline Service v2
• And manny Others

3. 3
Hadoop 3.0 Roadmap
• Hadoop 3.x Releases
• Planned for hadoop-3.0.0
Classpath isolation on by default HADOOP-11656
• hadoop-3.0.0-alpha1
– HADOOP
• Move to JDK8+
• Shell script rewrite HADOOP-9902
• Move default ports out of ephemeral range HDFS-9427
– HDFS
• Removal of hftp in favor of webhdfs HDFS-5570
• Support for more than two standby NameNodes HDFS-6440
• Support for Erasure Codes in HDFS HDFS-7285
• Intra-datanode balancer HDFS-1312
– YARN
• YARN Timeline Service v.2 YARN-2928
– MAPREDUCE
• Derive heap size or mapreduce.*.memory.mb automatically MAPREDUCE-5785

4. 4
Release Feature
2.0NameNode High-Availability,
2.2Federation, Snapshots, NFS v3 Mount
2.3Heterogenous Storage (Phase 1), In Memory Caching
2.4Rolling Upgrades, Posix ACL
2.5Extended Attributes
2.6Hot Swap Volumes, Heterogeneous Storage (Phase 2), transp. Encryption
2.7Files w/ variable Block Length, Inotify
2.8Docker Container in Linux, Yarn ATS 1.5, changing resources on alloc Yarn Container
2.9TBD
Release / Features / 2.X

5. 5
What's Apache Hadoop 3?
20142010 2011 201320122009 2015
2.2.0
Heterogeneous storage
HDFS in-memory caching
2.3.0 2.5.0
2.4.0
HDFS ACLs
2.0.0-alpha
2.1.0-beta
branch-1
(branch-0.20)
1.0.0 1.1.0 1.2.1(stable)0.20.1 0.20.205
0.22.0
Security
0.23.11(final)
NameNode Federation, YARN
0.21.0
New append
0.23.0
NameNode HA
branch-2
HDFS Snapshots
NFSv3 support
Windows
HDFS Rolling Upgrades
Application History Server
RM Automatic Failover
2.6.0
YARN Rolling Upgrades
Transparent Encryption
Archival Storage
2.7.0
Hadoop2
Drop JDK6 support
Truncate API
2016
branch-0.23
trunk
Hadoop3
Hadoop 3 and 2 were diverged in 2011 (5 years ago!)
Hadoop1 (EOL)
Source: Akira Ajisaka

6. 6
Break compatibility
Major version up is to clean up the code
• Deprecated APIs can be removed only in changing major version
– @Public and @Stable Java API
– REST API
– Metrics/JMX
– CLI
– Environment variables
• Wire-compatibility can be broken
– 2.X client cannot talk to 3.X server and vice versa
• Compatibility Guide:
– Apache Hadoop 3.0.0-alpha1 – Apache Hadoop Compatibility

7. 7
Erasure Coding

8. 8
Traditional Hadoop
• HDFS inherits 3-way replication from Google File System
• Simple, scalable and robust
• 200% storage overhead
• Secondary replicas rarely accessed

9. 9
Erasure Coding Saves Storage
• Simplified Example: storing 2 bits
• Same data durability
– can lose any 1 bit
• Pro: Half the storage overhead
• Cons: Slower recovery
1 01 0Replication: 2 extra bits
XOR Coding: 1 0⊕ 1= 1 extra bit

10. 10
Erasure Coding with m Data, n Parity
Reed Solomon Coding

11. 11
Durability and Efficiency
Data Durability = How many simultaneous failures can be tolerated?
Storage Efficiency = How much portion of storage is for useful data?
Data Durability Storage Efficiency
Single Replica 0 100%
3-way Replication 2 33%
XOR with 6 data cells 1 86%
RS (6,3) 3 67%
RS (10,4) 4 71%

12. 12
Continious Layout

13. 13
Stripped Layout

14. 14
Other Implementation

15. 15
HDFS Roadmap

16. 16
Erasure Coding: Current Status
 Phase 1: striping layout
 C = 64KB (default)
 Work for small files
 No data locality
 Available on trunk
 Phase 2: contiguous layout
 C = 128MB (= HDFS Block size)
 Not work for small files
 Data locality
 Now in progress (HDFS-8030)
Incoming Data
DataNode 1
DataNode 2
DataNode 3
DataNode 4
DataNode 5
・
・
・
Cell size (C)

17. 17
Name Server Changes
Mapping Logical and Storage Blocks
Too Many Storage Blocks?
Hierarchical Naming Protocol:

18. 18
Erasure Coding: Write files using (6,3)-Reed-Solomon
・
・
・
・
・
・
・
・
・
・
・
・
 Write data to 9 DNs in parallel
6 Data Blocks
DN1
DN6
DN7
Incoming Data
DN9
3 Parity Blocks
ECClient

19. 19
Erasure Coding: Read files
 Read data from 6 DNs in parallel
DN1
ECClient
DN6
DN9
・
・
・
・
・
・
・
・
・

20. 20
Erasure Coding: Read files when DN fails
 Read data from arbitrary 6 DNs in parallel
DN1
ECClient
×DN6
DN7
DN9
・
・
・
・
・
・
・
・
・

21. 21
hdfs erasure command
[loewe@loewe hadoop-3.0.0-alpha1]$ bin/hdfs erasurecode -help
Usage: hdfs erasurecode [generic options]
[-getPolicy <path>]
[-help [cmd ...]]
[-listPolicies]
[-setPolicy [-p <policyName>] <path>]
-getPolicy <path> :
Get erasure coding policy information about at specified path
-help [cmd ...] :
Displays help for given command or all commands if none is specified.
-listPolicies :
Get the list of erasure coding policies supported
-setPolicy [-p <policyName>] <path> :
Set a specified erasure coding policy to a directory
Options :
-p <policyName> erasure coding policy name to encode files. If not passed the
default policy will be used
<path> Path to a directory. Under this directory files will be
encoded using specified erasure coding policy

22. 22
Acceleration with Intel ISA-L
• 1 legacy coder
– From Facebook’s HDFS-RAID project
• 2 new coders
– Pure Java — code improvement over HDFS-RAID
– Native coder with Intel’s Intelligent Storage Acceleration Library (ISA-L)

23. 23
Benchmarks

24. 24
Benchmarks

25. 25
Benchmarks

26. 26
Benchmarks

27. 27
Yarn
Timeline Service v2

28. 28
First, A bit of Vision…
• Evolution of Hadoop start with YARN
• YARN Evolution will continue to drive Hadoop forward
• Hadoop 3 will still use Yarn, but there are a lot of Improvements
Hadoop 3

29. 29
Several important trends in age of Hadoop 3.0 +
YARN and Other Platform Services
Storage
Resource
Management Security
Service
Discovery Management
Monitoring
Alerts
IOT Assembly
Kafka Storm HBase Solr
Governance
MR Tez Spark …
Innovating
frameworks:
Flink,
DL(TensorFlow),
etc.
Various Environments
On Premise Private Cloud Public Cloud

30. 30
Yarn Architecture
Yarn
Resource
Database
Scheduler
Application
Timeline Service

31. 31
YARN Process Flow - Walkthrough
NodeManager NodeManager NodeManager NodeManager
Container
1.1
Container
2.4
NodeManager NodeManager NodeManager NodeManager
NodeManager NodeManager NodeManager NodeManager
Container
1.2
Container
1.3
AM 1
Container
2.2
Container
2.1
Container
2.3
AM2
Client2
Yarn
Scheduler
Yarn
Timeline Service
Client (Query)

32. 32
Why Timeline Service v2
• Scalability and reliability challenges
– Single instance of Timeline Server
– Storage (single local LevelDB instance)
• Usability
– Flow
– Metrics and configuration as first-class citizens
– Metrics aggregation up the entity hierarchy

33. 33
Highlights
v.1 v.2
Single writer/reader Timeline Server Distributed writer/collector architecture
Single local LevelDB storage* Scalable storage (HBase)
v.1 entity model New v.2 entity model
No aggregation Metrics aggregation
REST API Richer query REST API

34. 34
Architecture
• Separation of writers (“collectors”) and readers
• Distributed collectors: one collector for each app
• Dedicated RM collector for RM-generated data
• Collector discovery via RM
• Pluggable storage with HBase as default storage

35. 35
Distributed Collectors and Readers

36. 36
New Entity Model
• Flows and flow runs as parents of YARN applicaSon enSSes
• First-class configuraSon (key-value pairs)
• First-class metrics (single-value or Sme series)
• Designed to handle mulS-cluster environment out of the box

37. 37
What is a flow
• A flow is a group of YARNapplications
that are launched as parts of a logical
app
• Oozie, Scalding, Pig, etc.
– name:
– “frequent_visitor_stat”
– run id: 1466097809000
– version: “b9b9068”

38. 38
Metrics Aggregation
• Application level
– Rolls up sub-application metrics
– Performed in real time in the collectors
in memory
• Flow run level
– Rolls up app level metrics
– Performed in HBase region servers via
coprocessors
• Offline aggregation (TBD)
– Rolls up on user, queue, and flow offline
periodically
– Phoenix tables

39. 39
More Cloud Friendly
• Elastic
– Dynamic Resource Configuration
• YARN-291
• Allow tune down/up on NM’s resource in runtime
– Graceful decommissioning of NodeManagers
• YARN-914
• Drains a node that’s being decommissioned to allow running containers to finish
• Efficient
– Support for container resizing
• YARN-1197
• Allows applications to change the size of an existing container

40. 40
More Cloud Friendly (Contd.)
• Isolation
– Embrace container technology to achieve better isolation
– Resource isolation support for disk and network
• YARN-2619 (disk), YARN-2140 (network)
• Containers get a fair share of disk and network resources using Cgroups
– Docker support in LinuxContainerExecutor
• YARN-3611
• Support to launch Docker containers alongside process
• Packaging and resource isolation
• Operation
– Container upgrades (YARN-4726)
• ”Do an upgrade of my Spark / HBase apps with minimal impact to end-users”
– AM Restart With Work Preserving
• MAPREDUCE-6608

41. 41
• Add a native implementation of the map output collector
– Sort, Spill and IFile serialization
• Prequisites
– Built with -Pnative option
– Custom writable types and comparators are not supported
• Setting
<property name="mapreduce.job.map.output.collector.clas s"
value="org.apache.hadoop.mapred.nativetask.NativeMapOutputCollectorDelegator">
Task level native optimization (MAPREDUE-2841)

42. 42
Benchmark
• Release Note in the issue:
– "For shuffle-intensive jobs this may provide speed-ups of 30% or more."
• Benchmarked with 3 slaves (m3.xlarge)
– CentOS 7.2
– 3.0.0-SNAPSHOT (revision 5865fe2b)
• A very shuffle-intensive wordcount job
– Input: 2.6GB (compressed)
– Shuffle: 14GB
– Output: 10GB

43. 43
Updated Web UI

44. 44
Setup Timeline Service v2
• Set up the HBase cluster (1.1.x)
– Add the timeline service jar to HBase
– Install the flow run coprocessor
– Create tables via TimelineSchemaCreator utility
• Configure the YARN cluster
– Enable Timeline Service v.2
– Add hbase-site.xml for the timeline collector and readers
– Start the timeline reader daemon

45. 45
Shell Script rewrite

46. 46
Directory Structure

47. 47
Bin/hadoop Command

48. 48
bin/hdfs Command

49. 49
Shell Script Rewrite (HADOOP-9902)
• Hadoop and Shell Script
– Launching daemons
– Hadoop CLI
• Difficult to understand
– What is the correct env var to set a option
• java classpath?
• java.library.path?
• GC options?
– How to add the option to the env var
– We have to read almost all the shell scripts!
• New CLI is not completely downward compatible
– Hadoop 3: bin/hdfs namenode ‐format
– Hadoop 2: bin/hadoop namenode –format
Apache Hadoop 3.0.0-alpha1 – Apache Hadoop Compatibility

50. 50
After rewriting the scripts ...
• Easy to understand
• Because shell API doc is available
– Shelldoc maker generates docs
from the scripts
– Similar to JavaDoc
Public API
Documentation by Akira Ajisaka build (trunk):
http://aajisaka.github.io/hadoop-project/hadoop-project-dist/hadoop-common/UnixShellAPI.html

51. 51
• Very similar to .bashrc
– Read the API doc
– Create your own ~/.hadoopXX
• hadoop-env : hadoop-env.sh for each user
• hadooprc : called after shell env vars are configured
• And that's all :)
• ex.) Set additional classpath (.hadooprc)
hadoop_add_classpath /path/to/my/jar
.hadoop-env and .hadooprc
(HADOOP-11353, HADOOP-13045)

52. 52
--debug option is available
CLASSPATH was overwritten!!
(before HADOOP-13045)
Useful for troubleshooting
$ hadoop --debug version
DEBUG:
DEBUG:
DEBUG:
( s n i p )
DEBUG:
DEBUG:
DEBUG:
( s n i p )
DEBUG:
hadoop_parse_args: procesiong version
hadoop_parse: asking c a l l e r to skip 1
HADOOP_CONF_DIR=/usr/local/hadoop/ e t c / hadoop
Applying the u s e r ' s . hadooprc
I n i t i a l CLASSPATH=/path/to/my/jar
I n i t i a l i z e CLASSPATH
I n i t i a l CLASSPATH=/usr/ l o c a l / hadoop/ s h a r e / hadoop/common/lib/*

53. 53
Many new features, bug fixes, improvements
• 'hadoop distch' to change the ownership and permissions on many
files via MapReduce job
• 'hadoop jnipath' to print java.library.path
• 'hadoop --daemon' instead of hadoop-daemon.sh
– ex.) hdfs --daemon status namenode
– The return code for status is LSB-compatible
– hadoop-daemon(s).sh are now deprecated
• .out files are now appended (not overwritten)
– Allows external log rotation
• and many more
– see https://issues.apache.org/jira/browse/HADOOP-9902

54. 54
Derive heap size or mapreduce.*.memory.mb
automatically (MAPREDUCE-5785)
• In Hadoop 2, two similar properties must be set :(
– mapreduce.{map,reduce}.memory.mb
• The amount of memory to request from the scheduler for each task (ex. 2048)
– mapreduce.{map,reduce}.java.opts
• Java options for YARN containers (ex. -Xmx2G)
• In Hadoop 3, either is enough
– .java.opts is derived from .memory.mb and vice versa
• .java.opts = .memory.mb * mapreduce.job.heap.memory-mb.ratio
• .memory.mb = .java.opts / mapreduce.job.heap.memory-mb.ratio

55. 55
Intra Data-Node Balancer (HDFS-1312)
• Due to activities like deletes
the volumes of a DataNode
may become imbalance filled
DataNode
Block Placing
Policies

56. 56
Intra Data-Node Balancer (HDFS-1312)
• Offline scripts existed to reblance a DataNode
• HDFS-1312 introduces a online process that
rebalances the Volume of a DataNode
• „hdfs diskbalancer“ Command

57. 57
Multiple Name Nodes

58. 58
Support more than two NameNodes (HDFS-6440)
• Hadoop 2 now supports only 2 NameNodes
– 1 active and 1 standby
• Hadoop 3 supports 2 or more standby NameNodes
– provides additional fault-tolerance
– avoids multiple standby NNs to checkpoint at the same time
– # of standby should be small due to block report
(typically 3 or 5 NameNodes)

59. 59
Old Layout
ZK Failover
Controller
Active
NameNode
ZK Failover
Controller
Active
NameNode
Zookeeper Zookeeper Zookeeper
Fencing
Monitors the health of the NN Participating
in election of the active NN Coordinates
transition process Fence the other NN, if it
win election
The information which
NN is active is kept her

60. 60
New Layout
ZK Failover
Controller
Active
NameNode
ZK Failover
Controller
Standby
NameNode
Zookeeper Zookeeper Zookeeper
Fencing
Monitors the health of the NN Participating
in election of the active NN Coordinates
transition process Fence the other NN, if it
win election
The information which
NN is active is kept her
ZK Failover
Controller
Standby
NameNode
… 4 .. 5

61. 61
All Feature supported
• Checkpointing with NFS
• Checkpoint with Quorum Journal Daemon
• Manual Failover
• Automatic Failover

62. 62
Incompatible changes
• Many deprecated APIs will be removed
– hftp/hsftp/s3 -> webhdfs/s3{n,a}
– Metrics v1
– org.apache.hadoop.Records
– and more
• Improved CLI output
– 'mapred job -list' shows the job name as well
– 'hadoop fs -du' shows the raw disk usage, and aligned more unix-like
– and more
• Search 'Incompatible change' flag
– https://s.apache.org/sMO4

63. 63
Bump up the versions of the libraries
• Drop JDK7 support (HADOOP-11858)
• Dependency Hell
– Tomcat
– Jetty
– Jersey
– Guava
– Log4J
– Jackson
– And many more
common/lib/jackson-core-asl-1.9.13.jar:/usr/local/hadoop/share/hadoop/common/lib/commons-cli-1.2.jar:/usr/local/hadoop/share/hadoop/common/lib/apacheds-kerberos-codec-2.0.0-M15.jar:/usr/local/hadoop/share/
hadoop/common/lib/zookeeper-3.4.6.jar:/usr/local/hadoop/share/hadoop/common/lib/guava-11.0.2.jar:/usr/local/hadoop/share/hadoop/common/lib/api-asn1-api-1.0.0-M20.jar:/usr/local/hadoop/share/hadoop/
common/lib/slf4j-api-1.7.10.jar:/usr/local/hadoop/share/hadoop/common/lib/commons-beanutils-1.7.0.jar:/usr/local/hadoop/share/hadoop/common/lib/xmlenc-0.52.jar:/usr/local/hadoop/share/hadoop/common/lib/jsp-
api-2.1.jar:/usr/local/hadoop/share/hadoop/common/lib/curator-client-2.7.1.jar:/usr/local/hadoop/share/hadoop/common/lib/protobuf-java-2.5.0.jar:/usr/local/hadoop/share/hadoop/common/lib/jersey-json-1.9.jar:!/usr/
local/hadoop/share/hadoop/common/lib/jettison-1.1.jar:/usr/local/hadoop/share/hadoop/common/lib/commons-collections-3.2.2.jar:/usr/local/hadoop/share/hadoop/common/lib/commons-io-2.4.jar:/usr/local/hadoop/
share/hadoop/common/lib/jetty-util-6.1.26.jar:/usr/local/hadoop/share/hadoop/common/lib/nimbus-jose-jwt-3.9.jar:/usr/local/hadoop/share/hadoop/common/lib/commons-codec-1.4.jar:/usr/local/hadoop/share/
hadoop/common/lib/stax-api-1.0-2.jar:/usr/local/hadoop/share/hadoop/common/lib/junit-4.11.jar:/usr/local/hadoop/share/hadoop/common/lib/hamcrest-core-1.3.jar:/usr/local/hadoop/share/hadoop/common/lib/htrace-
core4-4.0.1-incubating.jar:/usr/local/hadoop/share/hadoop/common/lib/commons-beanutils-core-1.8.0.jar:/usr/local/hadoop/share/hadoop/common/lib/jersey-core-1.9.jar:/usr/local/hadoop/share/hadoop/common/lib/
netty-3.6.2.Final.jar:/usr/local/hadoop/share/hadoop/common/lib/hadoop-annotations-3.0.0-SNAPSHOT.jar:/usr/local/hadoop/share/hadoop/common/lib/jersey-server-1.9.jar:/usr/local! /hadoop/share/hadoop/
common/lib/activation-1.1.jar:/usr/local/hadoop/share/hadoop/common/lib/json-smart-1.1.1.jar:/usr/local/hadoop/share/hadoop/common/lib/apacheds-i18n-2.0.0-M15.jar:/usr/local/hadoop/share/hadoop/common/lib/
java-xmlbuilder-0.4.jar:/usr/local/hadoop/share/hadoop/common/lib/log4j-1.2.17.jar:/usr/local/hadoop/share/hadoop/common/lib/jsch-0.1.51.jar:/usr/local/hadoop/share/hadoop/common/lib/curator-
framework-2.7.1.jar:/usr/local/hadoop/share/hadoop/common/lib/httpcore-4.2.5.jar:/usr/local/hadoop/share/hadoop/common/lib/jcip-annotations-1.0.jar:/usr/local/hadoop/share/hadoop/common/lib/jaxb-api-2.2.2.jar:/
usr/local/hadoop/share/hadoop/common/lib/avro-1.7.4.jar:/usr/local/hadoop/share/hadoop/common/lib/commons-lang-2.6.jar:/usr/local/hadoop/share/hadoop/common/lib/httpclient-4.2.5.jar:/usr/local/hadoop/share/
hadoop/common/lib/hadoop-auth-3.0.0-SNAPSHOT.jar:/usr/local/hadoop/share/hadoop/common/lib/commons-digester-1.8.jar:/usr/local/hadoop/share/hadoop/common/lib/jsr305-3.0.0.jar:/usr/local/hadoop/share/
hadoop/common/lib/gson-2.2.4.jar:/usr/local/hadoop/share/hadoop/common/lib/jets3t-0.9.0.jar:/usr/local/hadoop/share/hadoop/common/lib/servlet-api-2.5.jar:/usr/local/hadoop/share/hadoop/common/lib/commons-
net-3.1.jar:/usr/local/hadoop/share/hadoop/common/lib/commons-math3-3.1.1.jar:/usr/local/hadoop/share/hadoop/common/lib/jackson-mapper-asl-1.9.13.jar:/usr/local/hadoop/share/hadoop/common/lib/commons-
compress-1.4.1.jar:/usr/local/hadoop/share/hadoop/common/lib/commons-logging-1.1.3.jar:/usr/local/hadoop/share/hadoop/common/lib/xz-1.0.jar:/usr/local/hadoop/share/hadoop/common/lib/mockito-all-1.8.5.jar:/
usr/local/hadoop/share/hadoop/common/lib/api-util-1.0.0-M20.jar:/usr/local/hadoop/share/hadoop/common/lib/commons-configuration-1.6.jar:/usr/local/hadoop/share/hadoop/common/lib/asm-3.2.jar:/usr/local/
hadoop/share/hadoop/common/lib/paranamer-2.3.jar:/usr/local/hadoop/share/hadoop/common/lib/jaxb-impl-2.2.3-1.jar:/usr/local/hadoop/share/hadoop/common/lib/snappy-java-1.0.4.1.jar:/usr/local/hadoop/share/
hadoop/common/lib/curator-recipes-2.7.1.jar:/usr/local/hadoop/share/hadoop/common/lib/jackson-xc-1.9.13.jar:/usr/local/hadoop/share/hadoop/common/lib/jackson-jaxrs-1.9.13.jar:/usr/local/hadoop/share/hadoop/
common/lib/slf4j-log4j12-1.7.10.jar:/usr/local/hadoop/share/hadoop/common/hadoop-common-3.0.0-SNAPSHOT.jar:/usr/local/hadoop/share/hadoop/common/hadoop-common-3.0.0-SNAPSHOT-tests.jar:/usr/local/
hadoop/share/hadoop/common/hadoop-nfs-3.0.0-SNAPSHOT.jar:/usr/local/hadoop/share/hadoop/hdfs:/usr/local/hadoop/share/hadoop/hdfs/lib/xml-apis-1.3.04.jar:/usr/local/hadoop/share/hadoop/hdfs/lib/hadoop-
hdfs-client-3.0.0-SNAPSHOT.jar:/usr/local/hadoop/share/hadoop/hdfs/lib/netty-all-4.1.0.Beta5.jar:/usr/local/hadoop/share/hadoop/hdfs/lib/hpack-0.11.0.jar:/usr/local/hadoop/share/hadoop/hdfs/lib/
xercesImpl-2.9.1.jar:/usr/local/hadoop/share/hadoop/hdfs/lib/leveldbjni-all-1.8.jar:/usr/local/hadoop/share/hadoop/hdfs/lib/commons-daemon-1.0.13.jar:/usr/local/hadoop/share/hadoop/hdfs/lib/okio-1.4.0.jar:/usr/local/
hadoop/share/hadoop/hdfs/lib/okhttp-2.4.0.jar:/usr/local/hadoop/share/hadoop/hdfs/hadoop-hdfs-nfs-3.0.0-SNAPSHOT.jar:/usr/local/hadoop/share/hadoop/hdfs/hadoop-hdfs-3.0.0-SNAPSHOT-tests.jar:/usr/local/
hadoop/share/hadoop/hdfs/hadoop-hdfs-3.0.0-SNAPSHOT.jar:/usr/local/hadoop/share/hadoop/mapreduce/hadoop-mapreduce-client-hs-3.0.0-SNAPSHOT.jar:/usr/local/hadoop/share/hadoop/mapreduce/hadoop-
mapreduce-client-jobclient-3.0.0-SNAPSHOT.jar:/usr/local/hadoop/share/hadoop/mapreduce/hadoop-mapreduce-client-nativetask-3.0.0-SNAPSHOT.jar:/usr/local/hadoop/share/hadoop/mapreduce/hadoop-
mapreduce-client-app-3.0.0-SNAPSHOT.jar:/usr/local/hadoop/share/hadoop/mapreduce/hadoop-mapreduce-client-core-3.0.0-SNAPSHOT.jar:/usr/local/hadoop/share/hadoop/mapreduce/hadoop-mapreduce-
examples-3.0.0-SNAPSHOT.jar:/usr/local/hadoop/share/hadoop/mapreduce/hadoop-mapreduce-client-hs-plugins-3.0.0-SNAPSHOT.jar:/usr/local/hadoop/share/hadoop/mapreduce/hadoop-mapreduce-client-
common-3.0.0-SNAPSHOT.jar:/usr/local/hadoop/share/hadoop/mapreduce/hadoop-mapreduce-client-jobclient-3.0.0-SNAPSHOT-tests.jar:/usr/local/hadoop/share/hadoop/mapreduce/hadoop-mapreduce-client-
shuffle-3.0.0-SNAPSHOT.jar:/usr/local/hadoop/share/hadoop/yarn/lib/javassist-3.18.1-GA.jar:/usr/local/hadoop/share/hadoop/yarn/lib/metrics-core-3.0.1.jar:/usr/local/hadoop/share/hadoop/yarn/lib/guice-3.0.jar:/usr/
local/hadoop/share/hadoop/yarn/lib/javax.inject-1.jar:/usr/local/hadoop/share/hadoop/yarn/lib/jersey-guice-1.9.jar:/usr/local/hadoop/share/hadoop/yarn/lib/aopalliance-1.0.jar:/usr/local/hadoop/share/hadoop/yarn/lib/
curator-test-2.7.1.jar:/usr/local/hadoop/share/hadoop/yarn/lib/fst-2.24.jar:/usr/local/hadoop/share/hadoop/yarn/lib/objenesis-2.1.jar:/usr/local/hadoop/share/hadoop/yarn/lib/jersey-client-1.9.jar:/usr/local/hadoop/
share/hadoop/yarn/lib/zookeeper-3.4.6-tests.jar:/usr/local/hadoop/share/hadoop/yarn/lib/guice-servle 0.jar:/usr/local/hadoop/share/hadoop/yarn/lib/commons-math-2.2.jar:/usr/local/hadoop/share/hadoop/yarn/
hadoop-yarn-server-applicationhistoryservice-3.0.0-SNAPSHOT.jar:/usr/local/hadoop/share/hadoop/yarn/hadoop-yarn-server-common-3.0.0-SNAPSHOT.jar:/usr/local/hadoop/share/hadoop/yarn/hadoop-yarn-
api-3.0.0-SNAPSHOT.jar:/usr/local/hadoop/share/hadoop/yarn/hadoop-yarn-server-tests-3.0.0-SNAPSHOT.jar:/usr/local/hadoop/share/hadoop/yarn/hadoop-yarn-common-3.0.0-SNAPSHOT.jar:/usr/local/hadoop/
share/hadoop/yarn/hadoop-yarn-server-resourcemanager-3.0.0-SNAPSHOT.jar:/usr/local/hadoop/share/hadoop/yarn/hadoop-yarn-applications-distributedshell-3.0.0-
SNAPSHOT.jar:/usr/local/hadoop/share/hadoop/ yarn/hadoop-yarn-server-web-proxy-3.0.0-SNAPSHOT.jar:/usr/local/hadoop/share/hadoop/yarn/hadoop-yarn-server-sharedcachemanager-3.0.0-
SNAPSHOT.jar:/usr/local/hadoop/share/hadoop/yarn/hadoop-yarn- registry-3.0.0-SNAPSHOT.jar:/usr/local/hadoop/share/hadoop/yarn/hadoop-yarn-server-nodemanager-3.0.0-
SNAPSHOT.jar:/usr/local/hadoop/share/hadoop/yarn/hadoop-yarn-server-timeline-pluginstorage-3.0.0-
SNAPSHOT.jar:/usr/local/hadoop/share/hadoop/yarn/hadoop-yarn-applications-unmanaged-am-launcher-3.0.0-SNAPSHOT.jar:/usr/local/hadoop/share/hadoop/yarn/hadoop-yarn-client-3.0.0-SNAPSHOT.jar:/usr/
java/latest/lib/tools.jar

64. 64
Classpath isolation (HADOOP-11656)
• Relaxing the "dependency hell"
– Separate client and server jars
– Client jar does not pull any third party dependencies
• If the isolation is done ...
– We can safely upgrade the libraries in server code
– In branch-2, the upgrade is incompatible :(

65. 65Follow me on Twitter: @loeweh