Robert Levas created AMBARI-17993:
-------------------------------------
Summary: Kerberos identity definitions in Kerberos descriptors
should explicitly declare a reference
Key: AMBARI-17993
URL: https://issues.apache.org/jira/browse/AMBARI-17993
Project: Ambari
Issue Type: Bug
Components: ambari-server
Affects Versions: 2.4.0
Reporter: Robert Levas
Assignee: Robert Levas
Priority: Blocker
Fix For: 2.4.0
Kerberos identity definitions in Kerberos descriptors should explicitly declare
a reference rather than rely on the identity's _name_ attribute.
Currently, the set of Kerberos identities declared at a service-level or a
component-level can contain identities with unique names. For example using:
{code}
"identities": [
{
"name": "identity",
"principal": {
"value": "service/_HOST@${realm}",
"configuration": "service-site/property1.principal",
...
},
"keytab": {
"file": "${keytab_dir}/service.service.keytab",
"configuration": "service-site/property1.keytab",
...
}
},
{
"name": "identity",
"principal": {
"value": "service/_HOST@${realm}",
"configuration": "service-site/property2.principal",
...
},
"keytab": {
"file": "${keytab_dir}/service.service.keytab",
"configuration": "service-site/property2.keytab",
...
}
}
]
{code}
Only the first "identity" principal is realized and the additional one is
ignored, leaving the configurations {service-site/property2.principal} and
{service-site/property2.keytab} untouched when Kerberos is enabled for the
service.
To help this, the 2nd instance can be converted to a reference, overriding only
the attributes the need to be changed - like the configurations.
{code}
"identities": [
{
"name": "identity",
"principal": {
"value": "service/_HOST@${realm}",
"configuration": "service-site/property1.principal",
...
},
"keytab": {
"file": "${keytab_dir}/service.service.keytab",
"configuration": "service-site/property1.keytab",
...
}
},
{
"name": "/SERVICE/identity",
"principal": {
"configuration": "service-site/property2.principal"
},
"keytab": {
"configuration": "service-site/property2.keytab"
}
}
]
{code}
This allows for both identity declarations to be realized, however this is
limited to only the 2 instances. If a 3rd instance is needed (to set an
additional configuration), it must look be:
{code}
{
"name": "/SERVICE/identity",
"principal": {
"configuration": "service-site/property3.principal"
},
"keytab": {
"configuration": "service-site/property3.keytab"
}
}
{code}
However since it's name is the same as the 2nd instance, it will be ignored.
If explicit references are specified, then multiple uniquely-named identity
blocks will be allowed to reference the same base identity, effectively
enabling the ability to declare unlimited configurations for the same identity
definition:
{code}
"identities": [
{
"name": "identity",
"principal": {
"value": "service/_HOST@${realm}",
"configuration": "service-site/property1.principal",
...
},
"keytab": {
"file": "${keytab_dir}/service.service.keytab",
"configuration": "service-site/property1.keytab",
...
}
},
{
"name": "identitiy_reference1",
"reference": "/SERVICE/identity",
"principal": {
"configuration": "service-site/property2.principal"
},
"keytab": {
"configuration": "service-site/property2.keytab"
}
},
{
"name": "identitiy_reference2",
"reference": "/SERVICE/identity",
"principal": {
"configuration": "service-site/property3.principal"
},
"keytab": {
"configuration": "service-site/property3.keytab"
}
}
]
{code}
NOTE: Backwards compatibility must be maintained when implementing this as to
not break existing Kerberos descriptors. So identity block names the look like
paths are to continue to be treated as references.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
