[
https://issues.apache.org/jira/browse/NIFI-13409?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Matthew Clarke updated NIFI-13409:
----------------------------------
Description:
Observed that NiFi is evaluating user identities against multiple configured
mapping patterns and applying the value and transforms for each matched pattern.
According to documentation, the expected behavior is that mapping are evaluated
in lexicographical order and first match only has its value and transform
applied.
Lets assume I have a node certificate with "CN=nifi-node1, ST=MD, C=US" and a
ldap username "nifiadmin" of the purposes of these observations and tests.
TEST 1:
Configured identity.mappings:
nifi.security.identity.mapping.pattern.dn=^CN=(.*?),\s{0,1}.+$
nifi.security.identity.mapping.transform.dn=NONE
nifi.security.identity.mapping.value.dn=$1
nifi.security.identity.mapping.pattern.ldap=^(.*)$
nifi.security.identity.mapping.transform.ldap=UPPER
nifi.security.identity.mapping.value.ldap=$1
Observations:
1. The nodes DN did have CN value extracted and also uppercased
NIFI-NODE1 <-- unexpected
2. My ldap users are transformed to all uppercase
NIFIAMDIN <-- expected
TEST 2:
Configured identity.mappings:
nifi.security.identity.mapping.pattern.dn=^CN=(.*?),\s{0,1}.+$
nifi.security.identity.mapping.transform.dn=NONE
nifi.security.identity.mapping.value.dn=$1
nifi.security.identity.mapping.pattern.ldap=^(.*)$
nifi.security.identity.mapping.transform.ldap=UPPER
nifi.security.identity.mapping.value.ldap=$1
nifi.security.identity.mapping.pattern.username=^(.*)$
nifi.security.identity.mapping.transform.username=LOWER
nifi.security.identity.mapping.value.username=$1
Observations:
1. The nodes DN did have CN value extracted and also uppercased
NIFI-NODE1 <-- unexpected
2. My ldap users are transformed to all uppercase
NIFIAMDIN <-- unexpected (assumed it would be lower if being stacked since
"username" mapping woudl be evaled after "ldap" mapping)
TEST 3:
Configured identity.mappings:
nifi.security.identity.mapping.pattern.dn=^CN=(.*?),\s{0,1}.+$
nifi.security.identity.mapping.transform.dn=NONE
nifi.security.identity.mapping.value.dn=$1
nifi.security.identity.mapping.pattern.ldap=^(.*)$
nifi.security.identity.mapping.transform.ldap=LOWER
nifi.security.identity.mapping.value.ldap=$1
nifi.security.identity.mapping.pattern.username=^(.*)$
nifi.security.identity.mapping.transform.username=UPPER
nifi.security.identity.mapping.value.username=$1
Observations:
1. The nodes DN did have CN value extracted and also uppercased
nifi-node1 <-- expected
2. My ldap users are transformed to all uppercase
nifiadmin <-- expected
Here everything remained lowercase as expected (node that both "ldap" and
"username" have same pattern regex set.
TEST 4:
Configured identity.mappings:
nifi.security.identity.mapping.pattern.dn=^CN=(.*?),\s{0,1}.+$
nifi.security.identity.mapping.transform.dn=NONE
nifi.security.identity.mapping.value.dn=$1
nifi.security.identity.mapping.pattern.ldap=^ldap(.*?)$
nifi.security.identity.mapping.transform.ldap=LOWER
nifi.security.identity.mapping.value.ldap=$1
nifi.security.identity.mapping.pattern.username=^(.*)$
nifi.security.identity.mapping.transform.username=UPPER
nifi.security.identity.mapping.value.username=$1
Observations:
1. The nodes DN did have CN value extracted and also uppercased
NIFI-NODE1 <-- unexpected
2. My ldap users are transformed to all uppercase
NIFIADMIN <-- expected
Appears that when multiple mappings have same pattern regex, only first is
having its value and transform stacked.
TEST 5:
Configured identity.mappings:
nifi.security.identity.mapping.pattern.dn=^CN=(.*?),\s{0,1}.+$
nifi.security.identity.mapping.transform.dn=NONE
nifi.security.identity.mapping.value.dn=$1
nifi.security.identity.mapping.pattern.ldap=^ldap(.*?)$
nifi.security.identity.mapping.transform.ldap=LOWER
nifi.security.identity.mapping.value.ldap=$1
nifi.security.identity.mapping.pattern.username=^(.*)$
nifi.security.identity.mapping.transform.username=UPPER
nifi.security.identity.mapping.value.username=$1.test
Observations:
this test was to see if only transforms were being stacked or both value and
transform were stacking.
1. The nodes DN did have CN value extracted and also uppercased
NIFI-NODE1.TEST <-- unexpected
2. My ldap users are transformed to all uppercase
NIFIADMIN.TEST <-- expected
This unexpected behavior causes issues when identities are coming from multiple
sources and user needs to uniquely manipulated those source identities. If one
pattern trims to a value that then matches another pattern, values and
transform stack.
Expectation is that identity mappings are evaluated by property name in
lexicographical order and only the first matching pattern as its corresponding
value and transform applied ignoring all other matching patterns.
was:
Observed that NiFi is evaluating user identities against multiple configured
mapping patterns and applying the value and transforms for each matched pattern.
According to documentation, the expected behavior is that mapping are evaluated
in lexicographical order and first match only has its value and transform
applied.
Lets assume I have a node certificate with "CN=nifi-node1, ST=MD, C=US" and a
ldap username "nifiadmin" of the purposes of these observations and tests.
TEST 1:
Configured identity.mappings:
nifi.security.identity.mapping.pattern.dn=^CN=(.*?),\s{0,1}.+$
nifi.security.identity.mapping.transform.dn=NONE
nifi.security.identity.mapping.value.dn=$1
nifi.security.identity.mapping.pattern.ldap=^(.*)$
nifi.security.identity.mapping.transform.ldap=UPPER
nifi.security.identity.mapping.value.ldap=$1
Observations:
1. The nodes DN did have CN value extracted and also uppercased
NIFI-NODE1 <-- unexpected
2. My ldap users are transformed to all uppercase
NIFIAMDIN <-- expected
TEST 2:
Configured identity.mappings:
nifi.security.identity.mapping.pattern.dn=^CN=(.*?),\s{0,1}.+$
nifi.security.identity.mapping.transform.dn=NONE
nifi.security.identity.mapping.value.dn=$1
nifi.security.identity.mapping.pattern.ldap=^(.*)$
nifi.security.identity.mapping.transform.ldap=UPPER
nifi.security.identity.mapping.value.ldap=$1
nifi.security.identity.mapping.pattern.username=^(.*)$
nifi.security.identity.mapping.transform.username=LOWER
nifi.security.identity.mapping.value.username=$1
Observations:
1. The nodes DN did have CN value extracted and also uppercased
NIFI-NODE1 <-- unexpected
2. My ldap users are transformed to all uppercase
NIFIAMDIN <-- unexpected (assumed it would be lower if being stacked since
"username" mapping woudl be evaled after "ldap" mapping)
TEST 3:
Configured identity.mappings:
nifi.security.identity.mapping.pattern.dn=^CN=(.*?),\s{0,1}.+$
nifi.security.identity.mapping.transform.dn=NONE
nifi.security.identity.mapping.value.dn=$1
nifi.security.identity.mapping.pattern.ldap=^(.*)$
nifi.security.identity.mapping.transform.ldap=LOWER
nifi.security.identity.mapping.value.ldap=$1
nifi.security.identity.mapping.pattern.username=^(.*)$
nifi.security.identity.mapping.transform.username=UPPER
nifi.security.identity.mapping.value.username=$1
Observations:
1. The nodes DN did have CN value extracted and also uppercased
nifi-node1 <-- expected
2. My ldap users are transformed to all uppercase
nifiadmin <-- expected
Here everything remained lowercase as expected (node that both "ldap" and
"username" have same pattern regex set.
TEST 4:
Configured identity.mappings:
nifi.security.identity.mapping.pattern.dn=^CN=(.*?),\s{0,1}.+$
nifi.security.identity.mapping.transform.dn=NONE
nifi.security.identity.mapping.value.dn=$1
nifi.security.identity.mapping.pattern.ldap=^ldap(.*?)$
nifi.security.identity.mapping.transform.ldap=LOWER
nifi.security.identity.mapping.value.ldap=$1
nifi.security.identity.mapping.pattern.username=^(.*)$
nifi.security.identity.mapping.transform.username=UPPER
nifi.security.identity.mapping.value.username=$1
Observations:
1. The nodes DN did have CN value extracted and also uppercased
NIFI-NODE1 <-- unexpected
2. My ldap users are transformed to all uppercase
NIFIADMIN <-- expected
Appears that when multiple mappings have same pattern regex, only first is
having its value and transform stacked.
TEST 5:
Configured identity.mappings:
nifi.security.identity.mapping.pattern.dn=^CN=(.*?),\s{0,1}.+$
nifi.security.identity.mapping.transform.dn=NONE
nifi.security.identity.mapping.value.dn=$1
nifi.security.identity.mapping.pattern.ldap=^ldap(.*?)$
nifi.security.identity.mapping.transform.ldap=LOWER
nifi.security.identity.mapping.value.ldap=$1
nifi.security.identity.mapping.pattern.username=^(.*).test$
nifi.security.identity.mapping.transform.username=UPPER
nifi.security.identity.mapping.value.username=$1
Observations:
this test was to see if only transforms were being stacked or both value and
transform were stacking.
1. The nodes DN did have CN value extracted and also uppercased
NIFI-NODE1.TEST <-- unexpected
2. My ldap users are transformed to all uppercase
NIFIADMIN.TEST <-- expected
This unexpected behavior causes issues when identities are coming from multiple
sources and user needs to uniquely manipulated those source identities. If one
pattern trims to a value that then matches another pattern, values and
transform stack.
Expectation is that identity mappings are evaluated by property name in
lexicographical order and only the first matching pattern as its corresponding
value and transform applied ignoring all other matching patterns.
> NiFi Identity mappings stacking rather then first match only being applied
> --------------------------------------------------------------------------
>
> Key: NIFI-13409
> URL: https://issues.apache.org/jira/browse/NIFI-13409
> Project: Apache NiFi
> Issue Type: Bug
> Affects Versions: 1.18.0, 1.23.1
> Reporter: Matthew Clarke
> Assignee: Simon Bence
> Priority: Critical
>
> Observed that NiFi is evaluating user identities against multiple configured
> mapping patterns and applying the value and transforms for each matched
> pattern.
> According to documentation, the expected behavior is that mapping are
> evaluated in lexicographical order and first match only has its value and
> transform applied.
> Lets assume I have a node certificate with "CN=nifi-node1, ST=MD, C=US" and a
> ldap username "nifiadmin" of the purposes of these observations and tests.
> TEST 1:
> Configured identity.mappings:
> nifi.security.identity.mapping.pattern.dn=^CN=(.*?),\s{0,1}.+$
> nifi.security.identity.mapping.transform.dn=NONE
> nifi.security.identity.mapping.value.dn=$1
> nifi.security.identity.mapping.pattern.ldap=^(.*)$
> nifi.security.identity.mapping.transform.ldap=UPPER
> nifi.security.identity.mapping.value.ldap=$1
> Observations:
> 1. The nodes DN did have CN value extracted and also uppercased
> NIFI-NODE1 <-- unexpected
> 2. My ldap users are transformed to all uppercase
> NIFIAMDIN <-- expected
> TEST 2:
> Configured identity.mappings:
> nifi.security.identity.mapping.pattern.dn=^CN=(.*?),\s{0,1}.+$
> nifi.security.identity.mapping.transform.dn=NONE
> nifi.security.identity.mapping.value.dn=$1
> nifi.security.identity.mapping.pattern.ldap=^(.*)$
> nifi.security.identity.mapping.transform.ldap=UPPER
> nifi.security.identity.mapping.value.ldap=$1
> nifi.security.identity.mapping.pattern.username=^(.*)$
> nifi.security.identity.mapping.transform.username=LOWER
> nifi.security.identity.mapping.value.username=$1
> Observations:
> 1. The nodes DN did have CN value extracted and also uppercased
> NIFI-NODE1 <-- unexpected
> 2. My ldap users are transformed to all uppercase
> NIFIAMDIN <-- unexpected (assumed it would be lower if being stacked since
> "username" mapping woudl be evaled after "ldap" mapping)
> TEST 3:
> Configured identity.mappings:
> nifi.security.identity.mapping.pattern.dn=^CN=(.*?),\s{0,1}.+$
> nifi.security.identity.mapping.transform.dn=NONE
> nifi.security.identity.mapping.value.dn=$1
> nifi.security.identity.mapping.pattern.ldap=^(.*)$
> nifi.security.identity.mapping.transform.ldap=LOWER
> nifi.security.identity.mapping.value.ldap=$1
> nifi.security.identity.mapping.pattern.username=^(.*)$
> nifi.security.identity.mapping.transform.username=UPPER
> nifi.security.identity.mapping.value.username=$1
> Observations:
> 1. The nodes DN did have CN value extracted and also uppercased
> nifi-node1 <-- expected
> 2. My ldap users are transformed to all uppercase
> nifiadmin <-- expected
> Here everything remained lowercase as expected (node that both "ldap" and
> "username" have same pattern regex set.
> TEST 4:
> Configured identity.mappings:
> nifi.security.identity.mapping.pattern.dn=^CN=(.*?),\s{0,1}.+$
> nifi.security.identity.mapping.transform.dn=NONE
> nifi.security.identity.mapping.value.dn=$1
> nifi.security.identity.mapping.pattern.ldap=^ldap(.*?)$
> nifi.security.identity.mapping.transform.ldap=LOWER
> nifi.security.identity.mapping.value.ldap=$1
> nifi.security.identity.mapping.pattern.username=^(.*)$
> nifi.security.identity.mapping.transform.username=UPPER
> nifi.security.identity.mapping.value.username=$1
> Observations:
> 1. The nodes DN did have CN value extracted and also uppercased
> NIFI-NODE1 <-- unexpected
> 2. My ldap users are transformed to all uppercase
> NIFIADMIN <-- expected
> Appears that when multiple mappings have same pattern regex, only first is
> having its value and transform stacked.
> TEST 5:
> Configured identity.mappings:
> nifi.security.identity.mapping.pattern.dn=^CN=(.*?),\s{0,1}.+$
> nifi.security.identity.mapping.transform.dn=NONE
> nifi.security.identity.mapping.value.dn=$1
> nifi.security.identity.mapping.pattern.ldap=^ldap(.*?)$
> nifi.security.identity.mapping.transform.ldap=LOWER
> nifi.security.identity.mapping.value.ldap=$1
> nifi.security.identity.mapping.pattern.username=^(.*)$
> nifi.security.identity.mapping.transform.username=UPPER
> nifi.security.identity.mapping.value.username=$1.test
> Observations:
> this test was to see if only transforms were being stacked or both value and
> transform were stacking.
> 1. The nodes DN did have CN value extracted and also uppercased
> NIFI-NODE1.TEST <-- unexpected
> 2. My ldap users are transformed to all uppercase
> NIFIADMIN.TEST <-- expected
> This unexpected behavior causes issues when identities are coming from
> multiple sources and user needs to uniquely manipulated those source
> identities. If one pattern trims to a value that then matches another
> pattern, values and transform stack.
> Expectation is that identity mappings are evaluated by property name in
> lexicographical order and only the first matching pattern as its
> corresponding value and transform applied ignoring all other matching
> patterns.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)