past, present and future of apache ambari

97
1 © Hortonworks Inc. 2011 – 2016. All Rights Reserved Apache Ambari Past, Present and Future September 2016

Upload: artem-ervits

Post on 21-Apr-2017

241 views

Category:

Data & Analytics


4 download

TRANSCRIPT

Page 1: Past, Present and Future of Apache Ambari

1   ©  Hortonworks  Inc.  2011  –  2016.  All  Rights  Reserved  

Apache  Ambari  Past,  Present  and  Future  September  2016  

Page 2: Past, Present and Future of Apache Ambari

2   ©  Hortonworks  Inc.  2011  –  2016.  All  Rights  Reserved  

Agenda  

Ã  Background,  Release  History  and  Demo  

Ã  Ambari  Alerts  and  Metrics  

Ã  Ambari  Blueprints  

Ã  Security  Setup  (Kerberos,  RBAC)  

Ã  Log  Search  (Technical  Preview)  

Ã  Automated  Cluster  Upgrade  

Ã  Extensibility:  Stacks  and  Views  

Ã  What’s  New  and  Next  

Page 3: Past, Present and Future of Apache Ambari

3   ©  Hortonworks  Inc.  2011  –  2016.  All  Rights  Reserved  

What  is  Apache  Ambari?  

A  completely  open  source  management  plaWorm  for  provisioning,  managing,  monitoring  and  securing  Apache  Hadoop  clusters.  Apache  Ambari  takes  the  guesswork  out  of  operaYng  Hadoop.  

Page 4: Past, Present and Future of Apache Ambari

4   ©  Hortonworks  Inc.  2011  –  2016.  All  Rights  Reserved  

What  Ambari  Does  

Simplified  InstallaYon,  ConfiguraYon  and  Management  

Centralized  Security  Setup  

Full  Visibility  into  Cluster  Health  

Highly  Extensible  and  Customizable  

•  Wizard-­‐driven  and  automated  cluster  provisioning  •  Smart  Configura>ons  and  Cluster  Recommenda>ons  •  Automated  Rolling  and  Express  cluster  upgrades  

•  Reduce  complexity  to  administer  security  across  the  plaWorm  •  Automate  setup  Kerberos  •  Simplify  the  configuraYon  of  Apache  Ranger  

•  Predefined  alerts  based  on  operaYonal  best  pracYces  •  Advanced  metrics  visualizaYon  with  Grafana  •  Integrated  with  SmartSense  for  proacYve  issues  prevenYon  

•  Seamlessly  fit  into  your  enterprise  environment  •  Bring  custom  Services  under  management  via  Ambari  Stacks  •  Customize  the  UI  with  Ambari  Views  

Page 5: Past, Present and Future of Apache Ambari

5   ©  Hortonworks  Inc.  2011  –  2016.  All  Rights  Reserved  

Ambari  Recent  Releases  

Ambari  2.2.0  Dec  2015  

Ambari  2.2.2  Apr  2016  

Ambari  2.4.0.0  GA  Aug  2016  

Page 6: Past, Present and Future of Apache Ambari

6   ©  Hortonworks  Inc.  2011  –  2016.  All  Rights  Reserved  

Ambari  Alerts  

Page 7: Past, Present and Future of Apache Ambari

7   ©  Hortonworks  Inc.  2011  –  2016.  All  Rights  Reserved  

Full  Visibility  into  Cluster  Health  

•  Centralized  management  of  Health  Alerts  

•  Pre-­‐defined  alerts,  configured  by  default  on  cluster  install  

Page 8: Past, Present and Future of Apache Ambari

8   ©  Hortonworks  Inc.  2011  –  2016.  All  Rights  Reserved  

Customizing  Alerts  

•  Control  thresholds,  check  intervals  and  response  text  

Page 9: Past, Present and Future of Apache Ambari

9   ©  Hortonworks  Inc.  2011  –  2016.  All  Rights  Reserved  

Alert  No>fica>ons  

•  What:  Create  and  manage  mul>ple  no>fica>on  targets  

•  Control  who  gets  noYfied  when  

•  Why:  Filter  by  severity  •  Send  only  certain  noYficaYons  to  certain  

targets  based  on  severity  

•  How:  Control  dispatch  method  •  Support  for  EMAIL  +  SNMP  

Page 10: Past, Present and Future of Apache Ambari

10   ©  Hortonworks  Inc.  2011  –  2016.  All  Rights  Reserved  

Alert  Groups  

•  Create  and  manage  groups  of  alerts  •  Group  alerts  further  controls  what  alerts  

are  dispatched  which  noYficaYons  

•  Assign  group  to  no>fica>ons  •  Only  dispatch  to  interested  parYes  

Page 11: Past, Present and Future of Apache Ambari

11   ©  Hortonworks  Inc.  2011  –  2016.  All  Rights  Reserved  

Alert  Check  Counts  

Ã  Customize  the  number  of  Ymes  an  alert  is  checked  before  dispatching  a  noYficaYon  

Ã  Avoid  dispatching  an  alert  noYficaYon  (email,  snmp)  in  case  of  transient  issues  

Page 12: Past, Present and Future of Apache Ambari

12   ©  Hortonworks  Inc.  2011  –  2016.  All  Rights  Reserved  

Configuring  the  Check  Count  

Ã  Set  globally  for  all  alerts,  or  override  for  a  specific  alert  

Global  Sefng  

Alert  Override  

Page 13: Past, Present and Future of Apache Ambari

13   ©  Hortonworks  Inc.  2011  –  2016.  All  Rights  Reserved  

State  Change  Types  

Ã  SOFT  state  changes  do  not  perform  a  dispatch  

Ã  HARD  state  changes  (to  non-­‐OK)  perform  dispatch  

Ã  Regardless  of  change:  –  The  Ambari  Web  UI  will  show  the  current  state  (OK/WARN/CRIT)  –  The  state  change  is  wrihen  to  ambari-alerts.log

2016-05-31 13:20:52,294 [CRITICAL] [SOFT] [AMBARI_METRICS] [grafana_webui] (Grafana Web UI) Connection failed to http://c6401.ambari.apache.org:3000 (<urlopen error [Errno 111] Connection refused>) 2016-05-31 13:22:52,290 [CRITICAL] [HARD] [AMBARI_METRICS] [grafana_webui] (Grafana Web UI) Connection failed to http://c6401.ambari.apache.org:3000 (<urlopen error [Errno 111] Connection refused>)

Note:  check  counts  are  not  configurable  for  AGGREGATE  alert  types.  All  state  changes  are  considered  HARD.  

Page 14: Past, Present and Future of Apache Ambari

14   ©  Hortonworks  Inc.  2011  –  2016.  All  Rights  Reserved  

Example:  Check  Count  =  3  

Check  1/3  State:  OK  

Change:  n/a  

Check  1/3  State:  OK  

Change:  n/a  

Check  1/3  State:  CRIT  

Change:  SOFT  

Check  2/3  State:  CRIT  Change:  n/a  

Check  3/3  State:  CRIT  

Change:  HARD  

Check  1/3  State:  OK  

Change:  HARD  

DISPATCH  

Check  Interval   Check  Interval   Check  Interval   Check  Interval   Check  Interval  

no  state  change  state  changes  to  CRIT  

performing  mulYple  checks  

back  to  OK  

Page 15: Past, Present and Future of Apache Ambari

15   ©  Hortonworks  Inc.  2011  –  2016.  All  Rights  Reserved  

Other  Items  

Ã  Alerts  Log  (AMBARI-­‐10249)  

•  Alert  state  changes  are  wrihen  to  /var/log/ambari-server/ambari-alerts.log

Ã  Script-­‐based  Alert  NoYficaYons  (AMBARI-­‐9919)  

•  Define  a  custom  script-­‐based  noYficaYon  dispatcher  

•  Executed  on  alert  state  changes  

2015-07-13 14:58:03,744 [OK] [ZOOKEEPER] [zookeeper_server_process] (ZooKeeper Server Process) TCP OK - 0.000s response on port 21812015-07-13 14:58:03,768 [OK] [HDFS] [datanode_process_percent] (Percent DataNodes Available) affected: [0], total: [1]

Page 16: Past, Present and Future of Apache Ambari

16   ©  Hortonworks  Inc.  2011  –  2016.  All  Rights  Reserved  

Ambari  Aler>ng  System  1.  User  creates  or  modifies  cluster  2.  Ambari  reads  alert  definiYons  from  

Stack  3.  Ambari  sends  alert  definiYons  to  

Agents  and  Agent  schedules  instance  checks  

4.  Agents  reports  alert  instance  status  in  the  heartbeat  

5.  Ambari  responds  to  alert  instance  status  changes  and  dispatches  noYficaYons  (if  applicable)  

Ambari Server

1  

2  

4  

Stack definition alerts.json

5  

Ambari Agent(s)

3  

emailsnmp

Page 17: Past, Present and Future of Apache Ambari

17   ©  Hortonworks  Inc.  2011  –  2016.  All  Rights  Reserved  

Alert  REST  APIs  

REST  Endpoint   Descrip>on  /api/v1/clusters/:clusterName/alert_definitions The  list  of  alert  definiYons  for  the  cluster.  

 /api/v1/clusters/:clusterName/alerts The  list  of  alert  instances  for  the  cluster.  Use  parYal  response  syntax  to  filter.  

 Example:  find  all  alert  instances  that  are  CRITITAL  /api/v1/clusters/c1/alerts?Alert/state.in(CRITICAL)  Example:  find  all  alert  instances  for  “ZooKeeper  Process”  alert  def  /api/v1/clusters/c1/alerts?Alert/definition_name=zookeeper_server_process

/api/v1/clusters/:clusterName/alert_groups The  list  of  alert  groups.    

/api/v1/clusters/:clusterName/alert_history The  list  of  alert  instance  status  changes.    

/api/v1/alert_targets/ The  list  of  configured  alert  noYficaYon  targets  for  Ambari.  

Page 18: Past, Present and Future of Apache Ambari

18   ©  Hortonworks  Inc.  2011  –  2016.  All  Rights  Reserved  

Ambari  Metrics  

Page 19: Past, Present and Future of Apache Ambari

19   ©  Hortonworks  Inc.  2011  –  2016.  All  Rights  Reserved  

Ambari  Metrics  

Ã  Built  using  Hadoop  technologies  

Ã  Three  components:  Metrics  Collector,  Metrics  Monitors,  Grafana  

Local Filesystem or HDFS

HBase ATS

Phoenix

Page 20: Past, Present and Future of Apache Ambari

20   ©  Hortonworks  Inc.  2011  –  2016.  All  Rights  Reserved  

Metrics  Collec>on  

1.  Metric  Monitors  send  system-­‐level  metrics  to  Collector  

2.  Sinks  send  Hadoop-­‐level  metrics  to  Collector  

3.  Metrics  Collector  service  stores  and  aggregates  metrics  

4.  Ambari  exposes  REST  API  for  metrics  retrieval  

Ambari Server

Metrics Monitor

Metrics Collector

Host1

Sink(s)

3  

Metrics Monitor

Host1

Sink(s) Metrics Monitor

Hosts

Sink(s)

1   2  

4  

Page 21: Past, Present and Future of Apache Ambari

21   ©  Hortonworks  Inc.  2011  –  2016.  All  Rights  Reserved  

Ambari  Metrics  +  Grafana:  Introduced  with  Ambari  2.2  

Ã  Including  Grafana  as  a  “NaYve  UI”  for  Ambari  Metrics  

Ã  Including  pre-­‐build  Dashboards  

Ã  Supports  configuring  for  HTTPS  

Ã  System  Home,  Servers  

Ã  HDFS  Home,  NameNodes,  DataNodes  

Ã  YARN  Home,  ApplicaYons,  Job  History  Server  

Ã  HBase  Home,  Performance,  Misc  

Highlights   List  of  Dashboards  

Page 22: Past, Present and Future of Apache Ambari

22   ©  Hortonworks  Inc.  2011  –  2016.  All  Rights  Reserved  

Grafana  includes  pre-­‐built  dashboards  for  visualizing  the  most  important  cluster  metrics.  

Page 23: Past, Present and Future of Apache Ambari

23   ©  Hortonworks  Inc.  2011  –  2016.  All  Rights  Reserved  

Ambari  Blueprints  

Page 24: Past, Present and Future of Apache Ambari

24   ©  Hortonworks  Inc.  2011  –  2016.  All  Rights  Reserved  

Ambari  Blueprints  

Ã  Two  primary  goals  of  Ambari  Blueprints:  

•  Provide  API-­‐driven  deployments  based  on  self-­‐contained  cluster  descripYon  

•  Ability  to  export  a  complete  cluster  descripYon  of  a  running  cluster  

Ã  Blueprints  contain  cluster  topology  and  configuraYon  informaYon  

Ã  Enables  interesYng  use  cases  between  physical  and  virtual  environments  

Page 25: Past, Present and Future of Apache Ambari

25   ©  Hortonworks  Inc.  2011  –  2016.  All  Rights  Reserved  

Ambari  Stacks  +  Blueprints  Together  

Stack  Defini>on  Component  Layout  &  

Configura>on  BLUEPRINT  

BLUEPRINT   INSTANTIATE   CLUSTER  HOSTS  

Page 26: Past, Present and Future of Apache Ambari

26   ©  Hortonworks  Inc.  2011  –  2016.  All  Rights  Reserved  

Blueprints  API  

BLUEPRINT POST /blueprints/my-blueprint

POST /clusters/MyCluster

1  

2  

CLUSTERCREATIONTEMPLATE

Page 27: Past, Present and Future of Apache Ambari

27   ©  Hortonworks  Inc.  2011  –  2016.  All  Rights  Reserved  

Example:  Single-­‐Node  Cluster  

{ "configurations" : [ { ”hdfs-site" : {

"dfs.namenode.name.dir" : ”/hadoop/nn" } } ], "host_groups" : [ { "name" : ”uber-host", "components" : [ { "name" : "NAMENODE” }, { "name" : "SECONDARY_NAMENODE” }, { "name" : "DATANODE” }, { "name" : "HDFS_CLIENT” }, { "name" : "RESOURCEMANAGER” }, { "name" : "NODEMANAGER” }, { "name" : "YARN_CLIENT” }, { "name" : "HISTORYSERVER” }, { "name" : "MAPREDUCE2_CLIENT” } ], "cardinality" : "1" } ], "Blueprints" : { "blueprint_name" : "single-node-hdfs-yarn", "stack_name" : "HDP", "stack_version" : "2.2" }}

{ "blueprint" : "single-node-hdfs-yarn", "host_groups" :[ { "name" : ”uber-host", "hosts" : [ { "fqdn" : "c6401.ambari.apache.org”

} ] } ]}

BLUEPRINT

CLUSTER CREATION TEMPLATE

Descrip>on  •  Single-­‐node  cluster  •  Use  HDP  2.2  Stack  •  HDFS  +  YARN  +  MR2  •  Everything  on  c6401  

Page 28: Past, Present and Future of Apache Ambari

28   ©  Hortonworks  Inc.  2011  –  2016.  All  Rights  Reserved  

Blueprints  Add  Host  

•  Add  hosts  to  a  cluster  based  on  a  host  group  from  a  Blueprint  

•  Add  one  or  more  hosts  with  a  single  call  

POST /api/v1/clusters/MyCluster/hosts{ "blueprint" : "myblueprint", "host_group" : "workers", "host_name" : "c6403.ambari.apache.org"}

POST /api/v1/clusters/MyCluster/hosts[ { "blueprint" : "myblueprint", "host_group" : "workers", "host_name" : "c6403.ambari.apache.org" }, { "blueprint" : "myblueprint", "host_group" : ”edge", "host_name" : "c6404.ambari.apache.org" }]

hhps://issues.apache.org/jira/browse/AMBARI-­‐8458  

Page 29: Past, Present and Future of Apache Ambari

29   ©  Hortonworks  Inc.  2011  –  2016.  All  Rights  Reserved  

Blueprints  Host  Discovery  (AMBARI-­‐10750)  

Ambari  

POST /api/v1/clusters/MyCluster/hosts

[ { "blueprint" : "single-node-hdfs-test2", "host_groups" :[ { "host_group" : "slave", "host_count" : 3, "host_predicate" : "Hosts/cpu_count>1” } ] }]

Page 30: Past, Present and Future of Apache Ambari

30   ©  Hortonworks  Inc.  2011  –  2016.  All  Rights  Reserved  

Where  Can  I  Learn  More  About  Blueprints?  

Ã  hhps://cwiki.apache.org/confluence/display/AMBARI/Blueprints  

Page 31: Past, Present and Future of Apache Ambari

31   ©  Hortonworks  Inc.  2011  –  2016.  All  Rights  Reserved  

Kerberos  Setup  

Page 32: Past, Present and Future of Apache Ambari

32   ©  Hortonworks  Inc.  2011  –  2016.  All  Rights  Reserved  

Background:  Hadoop  +  Kerberos  

Ã  Strongly  authenYcaYng  and  establishing  a  user’s  idenYty  is  the  basis  for  secure  access  in  Hadoop.  Users  need  to  be  able  to  reliably  “idenYfy”  themselves  and  then  have  that  idenYty  propagated  throughout  the  Hadoop  cluster.  

Ã  Once  this  is  done,  those  users  can  access  resources  (such  as  files  or  directories)  or  interact  with  the  cluster  (like  running  MapReduce  jobs).  

Ã  Besides  users,  Hadoop  cluster  resources  themselves  (such  as  Hosts  and  Services)  need  to  authenYcate  with  each  other  to  avoid  potenYal  malicious  systems  or  daemon’s  “posing  as”  trusted  components  of  the  cluster  to  gain  access  to  data.  

Page 33: Past, Present and Future of Apache Ambari

33   ©  Hortonworks  Inc.  2011  –  2016.  All  Rights  Reserved  

Background:  Hadoop  +  Kerberos  

Service  Component  

A  

Service  Component  

B  

Hadoop  Cluster  

KDC  

keytab  keytab  

Service  Component  

C  

keytab  

Service  Component  

D  

keytab  

Service  Component  

X  

Service  Component  

X  

keytab  keytab  

Service  Component  

X  

keytab  

Service  Component  

X  

keytab  

Kerberos  is  used  to  secure  the  

Components  in  the  cluster.  Kerberos  idenYYes  are  managed  via  

“keytabs”  on  the  Component  hosts.  

Principals  for  the  

cluster  are  managed  in  the  KDC.  

Page 34: Past, Present and Future of Apache Ambari

34   ©  Hortonworks  Inc.  2011  –  2016.  All  Rights  Reserved  

Automated  Kerberos  Security  Setup  

Wizard  Driven  Simplify  the  experience  of  enabling  and  managing  Kerberos  security  

Automated  and  Manual  Op>ons  CriYcal  for  today’s  enterprise  

Works  with  Exis>ng  Kerberos  Infrastructure  Such  as  AcYve  Directory  and  MIT  KDC  

Page 35: Past, Present and Future of Apache Ambari

35   ©  Hortonworks  Inc.  2011  –  2016.  All  Rights  Reserved  

Automated  vs.  Manual  Kerberos  Op>ons  

Automated   Manual  

KDC  Infrastructure   MIT,  AcYve  Directory   MIT,  AcYve  Directory,  FreeIPA  

Requires  KDC  administraYve  credenYals   Yes   No  

InstallaYon  of  Kerberos  clients   Yes,  opYonal   No  

Management  of  Kerberos  client  krb5.conf   Yes,  opYonal   No  

CreaYon  of  principals   Yes   No  

CreaYon  of  keytabs   Yes   No  

DistribuYon  of  keytabs   Yes   No  

Cluster  configuraYon   Yes   Yes  

Page 36: Past, Present and Future of Apache Ambari

36   ©  Hortonworks  Inc.  2011  –  2016.  All  Rights  Reserved  

Automated:  Principals  and  Keytabs  1.  User  provides  KDC  Admin  Account  

credenYals  to  Ambari  2.  Ambari  connects  to  KDC,  creates  

principals  (Service  and  Ambari)  needed  for  cluster  

3.  Ambari  generates  keytabs  for  the  principals  

4.  Ambari  distributes  keytabs  to  Ambari  Server  and  cluster  hosts  

5.  Ambari  discards  the  KDC  Admin  Account  credenYals  

Ambari Server KDC  

1   2  

4  

3  

5  

Cluster

Page 37: Past, Present and Future of Apache Ambari

37   ©  Hortonworks  Inc.  2011  –  2016.  All  Rights  Reserved  

Automated:  Customizable  Principal  Afributes  

•  Customize  the  directory  principal  ahributes  for  your  environment  

•  ParYcularly  useful  when  using  AcYve  Directory  for  Kerberos  

Page 38: Past, Present and Future of Apache Ambari

38   ©  Hortonworks  Inc.  2011  –  2016.  All  Rights  Reserved  

Automated:  Kerberos  Clients  

•  OpYon  to  manage  krb5.conf  

•  OpYon  to  install  clients  

OS   Client  RHEL/CentOS/OEL   krb5-­‐workstaYon  SLES  11   krb5-­‐client  Ubuntu  12   krb5-­‐user,  krb5-­‐

config  

Page 39: Past, Present and Future of Apache Ambari

39   ©  Hortonworks  Inc.  2011  –  2016.  All  Rights  Reserved  

Automated:  Kerberos  Opera>ons  

Ã  Once  enabled,  Ambari  works  directly  with  the  Kerberos  infrastructure  to  automate  these  common  tasks,  removing  the  burden  from  the  operator:  

•  Add/Delete  Host  

•  Add  Service  

•  Add/Delete  Component  

•  Regenerate  Keytabs  

•  Disable  Kerberos  

Page 40: Past, Present and Future of Apache Ambari

40   ©  Hortonworks  Inc.  2011  –  2016.  All  Rights  Reserved  

Manual:  Specify  Realm,  Client  U>li>es  Path  

Page 41: Past, Present and Future of Apache Ambari

41   ©  Hortonworks  Inc.  2011  –  2016.  All  Rights  Reserved  

Manual:  Principals  and  Keytabs  

Ã  Configure  IdenYYes  

Ã  Download  CSV  

Page 42: Past, Present and Future of Apache Ambari

42   ©  Hortonworks  Inc.  2011  –  2016.  All  Rights  Reserved  

Role-­‐Based  Access  Control  

Page 43: Past, Present and Future of Apache Ambari

43   ©  Hortonworks  Inc.  2011  –  2016.  All  Rights  Reserved  

Role-­‐Based  Access  Control  

Ã  Use  “roles”  for  more  granular  division  of  control  for  cluster  operaYons  

Old  Permission   New  Role   Notable  Permissions  

Operator   Cluster  Administrator   Full  operaYonal  control,  including  upgrades.  Ambari  Admins  are  implicitly  granted  this  Role.  

Cluster  Operator   Adding  and  removing  hosts.  

Service  Administrator   Manage  configuraYons,  move  components.  

Service  Operator   Service  stop  and  start  and  service-­‐specific  operaYons  such  as  HDFS  Rebalance.  

Read-­‐Only   Cluster  User   View  cluster  service  and  host  informaYon.  

Note:  Users  flagged  as  “Ambari  Admins”  are  implicitly  granted  Cluster  Administrator  permission.  

Page 44: Past, Present and Future of Apache Ambari

44   ©  Hortonworks  Inc.  2011  –  2016.  All  Rights  Reserved  

Managing  Cluster  Roles  Assign  roles  to  users  or  groups  

Manage  roles  in  Block  or  List  View  

layouts  

Page 45: Past, Present and Future of Apache Ambari

45   ©  Hortonworks  Inc.  2011  –  2016.  All  Rights  Reserved  

Managing  Cluster  Roles   View  users  or  groups  

Change  current  role  assignment  

Page 46: Past, Present and Future of Apache Ambari

46   ©  Hortonworks  Inc.  2011  –  2016.  All  Rights  Reserved  

Log  Search  

TECH  PREVIEW  

Page 47: Past, Present and Future of Apache Ambari

47   ©  Hortonworks  Inc.  2011  –  2016.  All  Rights  Reserved  

Log  Search  

Solr  

AMBAR I  Log  

Search  

Search  cluster  Component  Logs  from  within  Ambari!  

Goal:  When  issues  arise,  be  able  to  quickly  find  issues  across  all  cluster  components    ⬢  Capabili>es  

–  Rapid  Search  of  all  cluster  component  logs  –  Search  across  Yme  ranges,  log  levels,  and  for  keywords  

⬢  Core  Technologies:    –  Apache  Ambari  –  Apache  Solr  –  Apache  Ambari  Log  Search  

Page 48: Past, Present and Future of Apache Ambari

48   ©  Hortonworks  Inc.  2011  –  2016.  All  Rights  Reserved  

Log  Search  Architecture  

AMBAR I  

L O G    F E E D E R  

L O G    F E E D E R  

L O G  F E E D E R  

L O G  F E E D E R  

L O G    F E E D E R  

L O G    F E E D E R  

WOR K E R  NOD E  

WORK E R  NOD E  

WORK E R  NOD E  

WORK E R  NOD E  

WORK E R  NOD E  

WORK E R  NOD E  

Solr  

L OG    S E A R CH    

U I  

Page 49: Past, Present and Future of Apache Ambari

49   ©  Hortonworks  Inc.  2011  –  2016.  All  Rights  Reserved  

Log  Search  Details  

WORK E R  NOD E  

L O G    F E E D E R  

Solr  

L OG    S E A R CH    

U I  

Solr  

Solr  

AMBAR I  

Java  Process  MulY-­‐output  Support  Grok  

Solr  Cloud  Local  Disk  Storage  TTL  

Page 50: Past, Present and Future of Apache Ambari

50   ©  Hortonworks  Inc.  2011  –  2016.  All  Rights  Reserved  

Automated  Cluster  Upgrade  

Page 51: Past, Present and Future of Apache Ambari

51   ©  Hortonworks  Inc.  2011  –  2016.  All  Rights  Reserved  

Automated  Cluster  Upgrade  Terminology  

Rolling  Upgrade  Orchestrates  the  cluster  upgrade  in  an  order  that  is  meant  to  preserve  cluster  operaYon  and  minimize  service  impact.  The  tradeoff  is  more  stringent  prerequisites  and  a  longer  upgrade  Yme.  

Express  Upgrade  Orchestrates  the  cluster  upgrade  in  an  order  that  will  incur  cluster  downYme.  This  method  has  less  stringent  prerequisites  but  completes  in  a  faster  upgrade  Yme.  

Page 52: Past, Present and Future of Apache Ambari

52   ©  Hortonworks  Inc.  2011  –  2016.  All  Rights  Reserved  

Automated  Upgrade:  Rolling  or  Express  Choice  

Check  Prerequisites   Prepare   Register  +  

Install  Perform  Upgrade  

Perform  the  upgrade.  The  steps  depend  on  upgrade  method:  Rolling  or  Express.  

Finalize  

Finalize  the  upgrade,  making  the  target  version  the  current  

version.  

Perform  the  preparaYon  steps,  

which  include  making  backups  of  cri>cal  cluster  metadata.  

Review  the  prerequisites  to  

confirm  your  cluster  configuraYon  is  ready  

for  upgrade  

Register  the  sosware  repository  and  install  the  target  version  on  

the  cluster.  

Page 53: Past, Present and Future of Apache Ambari

53   ©  Hortonworks  Inc.  2011  –  2016.  All  Rights  Reserved  

Automated  Upgrade  Choice:  Rolling  or  Express  

Page 54: Past, Present and Future of Apache Ambari

54   ©  Hortonworks  Inc.  2011  –  2016.  All  Rights  Reserved  

Automated  Upgrade:  High-­‐Level  Process  

Register   Install   Perform  Upgrade   Finalize  

Page 55: Past, Present and Future of Apache Ambari

55   ©  Hortonworks  Inc.  2011  –  2016.  All  Rights  Reserved  

Register  New  Version  

•  Register  new  version  with  Ambari  

•  Provide  the  Repository  Base  URLs  for  new  version  

Register   Install   Perform  Upgrade   Finalize  

Page 56: Past, Present and Future of Apache Ambari

56   ©  Hortonworks  Inc.  2011  –  2016.  All  Rights  Reserved  

Install  Packages  

Ã  Click  “Install  Packages”  to  start  the  install  of  the  version  across  hosts  

Register   Install   Perform  Upgrade   Finalize  

Page 57: Past, Present and Future of Apache Ambari

57   ©  Hortonworks  Inc.  2011  –  2016.  All  Rights  Reserved  

Upgrade  Choice  

Ã  Upgrade  OpYons  dialog  gives  you  a  choice  of  method:  Rolling  or  Express  

Register   Install   Perform  Upgrade   Finalize  

Page 58: Past, Present and Future of Apache Ambari

58   ©  Hortonworks  Inc.  2011  –  2016.  All  Rights  Reserved  

Upgrade  Choice:  Rolling  or  Express  

Checks  indicate  if  an  opYon  is  available  

Register   Install   Perform  Upgrade   Finalize  

Page 59: Past, Present and Future of Apache Ambari

59   ©  Hortonworks  Inc.  2011  –  2016.  All  Rights  Reserved  

Upgrade  Choice:  Rolling  or  Express  

New  opYons  to  handle  failures    

Register   Install   Perform  Upgrade   Finalize  

Page 60: Past, Present and Future of Apache Ambari

60   ©  Hortonworks  Inc.  2011  –  2016.  All  Rights  Reserved  

Upgrade  Tolerance  Op>ons  

Tolerance  Op>on   Descrip>on  

Skip  all  Service  Check  failures   Ambari  will  automa>cally  skip  any  Service  Check  failures  and  complete  the  task  without  requiring  user  intervenYon  to  conYnue.  Aser  all  the  Service  Checks  have  run  in  a  task,  you  will  be  presented  with  summary  of  the  failures  and  an  opYon  to  conYnue  the  upgrade  or  pause.  

Skip  all  Slave  Component  failures   Ambari  will  automa>cally  skip  any  Slave  Component  failures  and  complete  the  task  of  upgrading  Slave  components  without  requiring  user  intervenYon  to  conYnue.  Aser  all  Slave  Components  have  been  upgraded,  you  will  be  presented  with  a  summary  of  the  failures  and  an  opYon  to  conYnue  the  upgrade  or  pause.  

Register   Install   Perform  Upgrade   Finalize  

Page 61: Past, Present and Future of Apache Ambari

61   ©  Hortonworks  Inc.  2011  –  2016.  All  Rights  Reserved  

Wizard  Driven  Experience  

With  verificaYon  and  validaYon  

Register   Install   Perform  Upgrade   Finalize  

Page 62: Past, Present and Future of Apache Ambari

62   ©  Hortonworks  Inc.  2011  –  2016.  All  Rights  Reserved  

Rolling  Upgrade  –  Success!  Register   Install   Perform  

Upgrade   Finalize  

Page 63: Past, Present and Future of Apache Ambari

63   ©  Hortonworks  Inc.  2011  –  2016.  All  Rights  Reserved  

Ambari  Extensibility  

Page 64: Past, Present and Future of Apache Ambari

64   ©  Hortonworks  Inc.  2011  –  2016.  All  Rights  Reserved  

Extensibility  Features  

•  To add new Services (ISV or otherwise) beyond Stack •  To customize a Stack for customer specific environments

•  To extend and customize the Ambari Web UI •  Add new capabilities, customize existing capabilities

Stacks

Views

Page 65: Past, Present and Future of Apache Ambari

65   ©  Hortonworks  Inc.  2011  –  2016.  All  Rights  Reserved  

Anatomy  of  Ambari  Extension  Points  

Ambari  Server  

Ambari  Agent  Ambari  

Agent  Ambari  Agent  

Ambari  Web  

Stacks  Stacks  

Stacks  

javajs python

Ambari  Views   Ambari  Stacks  

Page 66: Past, Present and Future of Apache Ambari

66   ©  Hortonworks  Inc.  2011  –  2016.  All  Rights  Reserved  

Ambari  Stacks  

Page 67: Past, Present and Future of Apache Ambari

67   ©  Hortonworks  Inc.  2011  –  2016.  All  Rights  Reserved  

Ambari  Stacks  

Ã  Defines  a  consistent  Stack  lifecycle  interface  that  can  be  extended  

Ã  Encapsulates  Stack  Versions,  Services,  Components,  Dependencies,  Cardinality,  ConfiguraYons,  Commands    

Ã  Dynamically  add  Stack  +  Service  definiYons  

AMBARI  {rest}

<ambari-web>

Stacks  

HDFS   YARN   MR2  

Hive  

Pig  

Oozie  HBase  

Storm  Falcon  

Page 68: Past, Present and Future of Apache Ambari

68   ©  Hortonworks  Inc.  2011  –  2016.  All  Rights  Reserved  

Stack  Terminology  Term   Defini>on   Examples  

STACK   Defines  a  set  of  Services,  where  to  obtain  the  sosware  packages  and  how  to  manage  the  lifecycle.   HDP-­‐2.4,  HDP-­‐2.5  

SERVICE   Defines  the  Components  that  make-­‐up  the  Service.   HDFS,  YARN,  HIVE,  SPARK  

COMPONENT   The  building-­‐blocks  of  a  Service,  that  adhere  to  a  certain  lifecycle.     NAMENODE,  DATANODE,  OOZIE_SERVER  

CATEGORY   The  category  of  Component.   MASTER,  SLAVE,  CLIENT  

Page 69: Past, Present and Future of Apache Ambari

69   ©  Hortonworks  Inc.  2011  –  2016.  All  Rights  Reserved  

Stack  Mechanics  

Ã  Stacks  define  Services  +  Repos  –  What  is  in  the  Stack,  and  where  to  get  the  bits  

Ã  Each  Service  has  a  definiYon  –  What  Components  are  part  of  the  Service  

Ã  Each  Service  has  defined  lifecycle  commands  –  start,  stop,  status,  install,  configure  

Ã  Lifecycle  is  controlled  via  command  scripts  

Ã  Ability  to  define  “custom”  commands  

AMBARI  SERVER  

Stack  

Command  Scripts  

Service  Defini>ons  

AMBARI  AGENT/S  AMBARI  AGENT/S  AMBARI  AGENT/S  

pythonxml

Repos  

Page 70: Past, Present and Future of Apache Ambari

70   ©  Hortonworks  Inc.  2011  –  2016.  All  Rights  Reserved  

Stacks  Support  Inheritance  

HDP  2.0  Stack  

HDP  2.4  Stack  

•  Defines  a  set  of  Service  definiYons  •  Default  service  configuraYons  and  command  scripts  

•  Overrides  any  Service  definiYons,  commands  and  configuraYons  

•  Adds  new  Services  specific  to  this  Stack  

Page 71: Past, Present and Future of Apache Ambari

71   ©  Hortonworks  Inc.  2011  –  2016.  All  Rights  Reserved  

Where  Can  I  Learn  More  About  Stacks?  

Ã  hhps://cwiki.apache.org/confluence/pages/viewpage.acYon?pageId=38571133  

Ã  hhps://github.com/apache/ambari/tree/trunk/ambari-­‐server/src/main/resources/stacks/HDP/  

Page 72: Past, Present and Future of Apache Ambari

72   ©  Hortonworks  Inc.  2011  –  2016.  All  Rights  Reserved  

Guided  Configs  

Page 73: Past, Present and Future of Apache Ambari

73   ©  Hortonworks  Inc.  2011  –  2016.  All  Rights  Reserved  

Guided  Configura>ons  

Ã  Take  the  guesswork  out  of  cluster  configuraYon  

Ã  IntuiYve  layout  and  grouping  of  configuraYons  

Ã  Smart  UI  controls  to  make  it  easier  to  set  values  

Ã  RecommendaYons  and  cross-­‐service  dependency  checks  

Ã  Take  the  guesswork  out  of  cluster  configuraYon  

Ã  Driven  by  Stack  definiYon  

Page 74: Past, Present and Future of Apache Ambari

74   ©  Hortonworks  Inc.  2011  –  2016.  All  Rights  Reserved  

Layout  and  Grouping  

Subtabs  

Page 75: Past, Present and Future of Apache Ambari

75   ©  Hortonworks  Inc.  2011  –  2016.  All  Rights  Reserved  

UI  Controls  

Recommended  Bounds  

Recommended  Value  

Set  Recommended  

Escape  Hatch  

Page 76: Past, Present and Future of Apache Ambari

76   ©  Hortonworks  Inc.  2011  –  2016.  All  Rights  Reserved  

Control  “Escape  Hatch”  

Page 77: Past, Present and Future of Apache Ambari

77   ©  Hortonworks  Inc.  2011  –  2016.  All  Rights  Reserved  

Dependency  Checking  

Page 78: Past, Present and Future of Apache Ambari

78   ©  Hortonworks  Inc.  2011  –  2016.  All  Rights  Reserved  

Driven  By  Stack  Defini>on:  Themes  

Page 79: Past, Present and Future of Apache Ambari

79   ©  Hortonworks  Inc.  2011  –  2016.  All  Rights  Reserved  

Tabs  

Tab1   Tab2  

Page 80: Past, Present and Future of Apache Ambari

80   ©  Hortonworks  Inc.  2011  –  2016.  All  Rights  Reserved  

Sec>ons  

Sec>on1   Sec>on2  

Page 81: Past, Present and Future of Apache Ambari

81   ©  Hortonworks  Inc.  2011  –  2016.  All  Rights  Reserved  

SubSec>ons  

SubSec>on1   SubSec>on2   SubSec>on1  

Page 82: Past, Present and Future of Apache Ambari

82   ©  Hortonworks  Inc.  2011  –  2016.  All  Rights  Reserved  

UI  Controls  and  Placement  

Control  +  Placement  

Control  +  Placement  

Control  +  Placement   Control  +  Placement  

Control  +  Placement  

Page 83: Past, Present and Future of Apache Ambari

83   ©  Hortonworks  Inc.  2011  –  2016.  All  Rights  Reserved  

Ambari  Views  

Page 84: Past, Present and Future of Apache Ambari

84   ©  Hortonworks  Inc.  2011  –  2016.  All  Rights  Reserved  

Ambari  Views  Framework  

Ã  Developers  can  extend  the  Ambari  Web  interface  •  Views  expose  custom  UI  features  for  Hadoop  Services  

Ã  Ambari  Admins  can  enYtle  Views  to  Ambari  Web  users  •  EnYtlements  framework  for  controlling  access  to  Views  

Page 85: Past, Present and Future of Apache Ambari

85   ©  Hortonworks  Inc.  2011  –  2016.  All  Rights  Reserved  

Views  Framework  

Views  Framework  vs.  Views  

Views  

Core  to  Ambari  

Built  by  Hortonworks,  Community,  Partners  

Page 86: Past, Present and Future of Apache Ambari

86   ©  Hortonworks  Inc.  2011  –  2016.  All  Rights  Reserved  

Views  Framework  

Views  Framework  vs.  Views  

Views  

Core  to  Ambari  

Built  by  Hortonworks,  Community,  Partners  

Page 87: Past, Present and Future of Apache Ambari

87   ©  Hortonworks  Inc.  2011  –  2016.  All  Rights  Reserved  

Example  Views  

Capacity  Scheduler  “Queue  Manager”  View  

Tez  “Jobs”  View  

Page 88: Past, Present and Future of Apache Ambari

88   ©  Hortonworks  Inc.  2011  –  2016.  All  Rights  Reserved  

View  Components  

Ã  Serve  client-­‐side  assets  (such  as  HTML  +  JavaScript)  

Ã  Expose  server-­‐side  resources  (such  as  REST  endpoints)  

VIEW  Client-­‐side  assets  

(.js,  html)  

AMBARI  WEB  

VIEW  Server-­‐side  resources  (java)  

AMBARI  SERVER  

{rest}

Hadoop  and  other  

systems  

Page 89: Past, Present and Future of Apache Ambari

89   ©  Hortonworks  Inc.  2011  –  2016.  All  Rights  Reserved  

View  Delivery  

1.  Develop  the  View  (just  like  you  would  for  a  Web  App)  

2.  Package  as  a  View  (basically  a  WAR)  

3.  Deploy  the  View  into  Ambari  

4.  Ambari  Admins  create  +  configuraYon  view  instance(s)  and  give  access  to  users  +  groups  

Develop   Deploy  Package  Create  

Instance(s)  

Page 90: Past, Present and Future of Apache Ambari

90   ©  Hortonworks  Inc.  2011  –  2016.  All  Rights  Reserved  

Versions  and  Instances  

Ã  Deploy  mulYple  versions  and  create  mulYple  instances  of  a  view  

Ã  Manage  accessibility  and  usage  

Page 91: Past, Present and Future of Apache Ambari

91   ©  Hortonworks  Inc.  2011  –  2016.  All  Rights  Reserved  

Choice  of  Deployment  Model  

Ã  For  Hadoop  Operators:  –  Deploy  Views  in  an  Ambari  Server  that  is  managing  a  Hadoop  cluster  

Ã  For  Data  Workers:  –  Run  Views  in  a  “standalone”  Ambari  Server  

Ambari Server

HADOOP  Store  &  Process  

Ambari Server

Operators  manage  the  cluster,  may  have  Views  deployed  

Data  Workers  use  the  cluster  and  use  a  “standalone”  Ambari  Server  for  Views  

   

Page 92: Past, Present and Future of Apache Ambari

92   ©  Hortonworks  Inc.  2011  –  2016.  All  Rights  Reserved  

Where  Can  I  Learn  More  about  Views?  

Ã  hhps://cwiki.apache.org/confluence/display/AMBARI/Views  

Ã  hhps://github.com/apache/ambari/blob/trunk/ambari-­‐views/docs/index.md  

Ã  hhps://github.com/apache/ambari/tree/trunk/ambari-­‐views/examples  

Ã  hhps://github.com/apache/ambari/tree/trunk/contrib/views  

Page 93: Past, Present and Future of Apache Ambari

93   ©  Hortonworks  Inc.  2011  –  2016.  All  Rights  Reserved  

What’s  New  and  Next  

Page 94: Past, Present and Future of Apache Ambari

94   ©  Hortonworks  Inc.  2011  –  2016.  All  Rights  Reserved  

What’s  New  in  Ambari  2.4  

Ã  Alerts:  Retry  tolerance  (AMBARI-­‐15686)  

Ã  Alerts:  Customizable  SCRIPT  Parameters  (AMBARI-­‐14898)  

Ã  Alerts:  New  HDFS  Alerts  (AMBARI-­‐14800)  

Ã  New  Host  Page  Filtering  (AMBARI-­‐15210)  

Ã  Remove  Service  (AMBARI-­‐14759)  

Ã  Support  for  SLES  12  Technical  Preview  (AMBARI-­‐16007)  

Ã  Stability:  Database  Consistency  Checking  (AMBARI-­‐16258)  

Ã  Customizable  Ambari  Log  +  PID  Dirs  (AMBARI-­‐15300)  

Ã  New  Version  RegistraYon  Experience  (AMBARI-­‐15724)  

Ã  Log  Search  Technical  Preview  (AMBARI-­‐14927)  

Ã  OperaYonal  Audit  Logging  (AMBARI-­‐15241)  

Ã  Role-­‐Based  Access  Control  (AMBARI-­‐13977)  

Ã  Automated  Setup  of  Ambari  Kerberos  (AMBARI-­‐15561)  

Ã  Automated  Setup  of  Ambari  Proxy  User  (AMBARI-­‐15561)  

Ã  Customizable  Host  Reg.  SSH  Port  (AMBARI-­‐13450)  

Core  Features   Security  Features  

Ã  View  URLs  (AMBARI-­‐15821),  View  Refresh  (AMBARI-­‐15682)  

Ã  Inherit  Cluster  Permissions  (AMBARI-­‐16177)  

Ã  Remote  Cluster  RegistraYon  (AMBARI-­‐16274)  

Views  Framework  Features  

Page 95: Past, Present and Future of Apache Ambari

95   ©  Hortonworks  Inc.  2011  –  2016.  All  Rights  Reserved  

What’s  Next  

Ã  Log  Search  GA  (AMBARI-­‐14927)  

Ã  Service  and  Component  “auto-­‐start”  (AMBARI-­‐2330)  

Ã  Ambari  Server  HA  (AcYve-­‐Passive  +  Load  Balancer)  (AMBARI-­‐17126)  

Ã  MulY-­‐service  instances  &  versions  (AMBARI-­‐14714)  

Page 96: Past, Present and Future of Apache Ambari

96   ©  Hortonworks  Inc.  2011  –  2016.  All  Rights  Reserved  

Learn  More  About  Apache  Ambari  

Resource   Loca>on  

Apache  Ambari  Project  Page   hhp://ambari.apache.org  

Ambari  Project  Wiki   hhps://cwiki.apache.org/confluence/display/AMBARI  

Ambari  Project  JIRA   hhps://issues.apache.org/jira/browse/AMBARI  

Stacks   hhps://cwiki.apache.org/confluence/pages/viewpage.acYon?pageId=38571133    

Blueprints   hhps://cwiki.apache.org/confluence/display/AMBARI/Blueprints    

Views   hhps://cwiki.apache.org/confluence/display/AMBARI/Views    

Page 97: Past, Present and Future of Apache Ambari

97   ©  Hortonworks  Inc.  2011  –  2016.  All  Rights  Reserved  

Thank  You