[AMBARI-21868] Implement host recovery - backend changes

• Type: Task
• Status: Resolved
• Priority: Blocker
• Resolution: Fixed
• Affects Version/s: 2.6.0
• Fix Version/s: 2.6.0
• Component/s: ambari-server
• Assignee: Eugene Chekanskiy
• Reporter: Dmytro Sen
• Created: 01/Sep/17 06:43
• Updated: 08/Sep/17 21:34
• Resolved: 08/Sep/17 20:54

Description

Problem Statement:
Allow user to bring back a host/VM that is dead with the same hostname and allow an action to recover the state of the host to what it was before the incident

Solution:

• User gets the host/VM back with same hostname (different IP address is ok)
• The UI state would be INSTALLED for all host components as soon as the host starts heart beating
• User will initiate a Recover user action that will do the following (batch operation):
  • Move the state of the components back to INIT (allow this state transition)

    curl -H "X-Requested-By:ambari" -u admin:admin -i -X PUT -d {"RequestInfo":{"context":"Reset All Host Components","operation_level":{"level":"HOST","cluster_name":"c1","host_names":"c6401.ambari.apahe.org"},"query":"HostRoles/component_name.in(DATANODE,NODEMANAGER,METRICS_MONITOR)"},"Body":{"HostRoles":{"state":"INIT"}}} http://c6401.ambari.apache.org:8080/api/v1/clusters/c1/hosts/c6401.ambari.apache.org/host_components
    

  • KERBEROS only steps:
    • Issue an INSTALL operation on KERBEROS_CLIENT hostcomponent
    • Issue PUT request

      PUT http://server:8080/api/v1/clusters/cl1?regenerate_keytabs=all&regenerate_hosts=recover_hostname1,recover_hostname2,...&ignore_config_updates=true
      
      {"Clusters":{"security_type":"KERBEROS"}}
      

  • Issue an INSTALL operation on all hostcomponents (Same call as above with state = "INSTALLED")

Issue Links

• links to: review