Loading...

XML

Word

Printable

Details

Type: Bug
Resolution: Done
Priority: Major
Fix Version/s: Logging 5.7.3
Affects Version/s: Logging 5.6.z, Logging 5.7.z
Component/s: Log Collection
Labels:
- devel_ack+

Blocked:
False
Blocked Reason:
None
Ready:
False
Docs QE Status:
NEW
QE Status:
NEW
Release Note Text:

Hide
Before this update, Fluentd is unable to send logs to an Elasticsearch cluster when a passphrase is applied to the private key provided.
With this chages, the issue was resolved by adding the necessary configuration to the Fluentd Elasticsearch plugin, Fluentd now properly handles passphrase-protected private keys when establishing a connection with Elasticsearch.

Show
Before this update, Fluentd is unable to send logs to an Elasticsearch cluster when a passphrase is applied to the private key provided. With this chages, the issue was resolved by adding the necessary configuration to the Fluentd Elasticsearch plugin, Fluentd now properly handles passphrase-protected private keys when establishing a connection with Elasticsearch.
Release Note Type:
Bug Fix

Sprint:
Log Collection - Sprint 238
Severity:
Important

Test Link:
https://polarion.engineering.redhat.com/polarion/#/project/OSE/workitem?id=OCP-65036

SFDC Cases Links:
SFDC Cases Counter:

Description

Description of problem:

The clusterlogging operator is not generating the right configuration missing to generate the {}client_key_pass{} for fluentd when the output defined is Elasticsearch and using passphrase

Version-Release number of selected component (if applicable):

/// Logging version$ ns="openshift-logging"$ oc get csv -n $nsNAME                                       DISPLAY                                          VERSION                   REPLACES                        PHASEcluster-logging.v5.7.2                     Red Hat OpenShift Logging                        5.7.2                     cluster-logging.v5.7.1          Succeededelasticsearch-operator.v5.7.2              OpenShift Elasticsearch Operator                 5.7.2                     elasticsearch-operator.v5.7.1   Succeeded

Also tested and failing in latest from stable-5.6

How reproducible:

Always

Steps to Reproduce:

/// clusterlogging configuration running fluentd
$ oc get clusterlogging instance -n $ns -o yaml
...
apiVersion: "logging.openshift.io/v1"
kind: "ClusterLogging"
metadata:
  name: "instance"
  namespace: "openshift-logging"
spec:
  collection:
    logs:
      type: "fluentd"
      fluentd:
        resources: {}


/// Create secret containing certificates + passphrase for private key
$ oc create secret generic -n openshift-logging es-secret --from-file=tls.key=tls.key --from-file=tls.crt=tls.crt --from-file=ca-bundle.crt=ca-bundle.crt --from-literal=username=user --from-literal=password=pass --from-literal=passphrase=test

/// clusterlogforwarder configured to send to an elasticsearch output using certificates that requires passphrase
$ oc get clusterlogforwarder instance -o yaml -n $ns
apiVersion: logging.openshift.io/v1
kind: ClusterLogForwarder
metadata:
  creationTimestamp: "2023-06-23T10:25:47Z"
  generation: 3
  name: instance
  namespace: openshift-logging
  resourceVersion: "2863397"
  uid: 7bc98f0e-c985-4e0b-bbfb-2dc0b5620b6f
spec:
  outputs:
  - elasticsearch:
      version: 8
    name: test
    secret:
      name: es-secret
    type: elasticsearch
    url: https://elasticsearch:9200
  pipelines:
  - inputRefs:
    - infrastructure
    - application
    - audit
    name: all-to-default
    outputRefs:
    - test

Actual results:

Collector not able to establish SSL connection because not able to open the private key and returning the error:

/// Review is present the error "Enter PEM pass phrase
$ for pod in $(oc get pods -l component=collector -o name -n $ns); do oc logs $pod -n $ns -c collector |grep -i "Enter PEM"; done|head -1
Enter PEM pass phrase:

The fluentd configuration generated by the clusterlogging operator doesn't contain the {}client_key_pass{} :

/// Confirm in the fluentd configuration that is as below 
$ 

  <match **>
    @type elasticsearch
    @id test
    host elasticsearch
    port 9200
    scheme https
    ssl_version TLSv1_2
    user "#{File.exists?('/var/run/ocp-collector/secrets/es-secret/username') ? open('/var/run/ocp-collector/secrets/es-secret/username','r') do |f|f.read end : ''}"
    password "#{File.exists?('/var/run/ocp-collector/secrets/es-secret/password') ? open('/var/run/ocp-collector/secrets/es-secret/password','r') do |f|f.read end : ''}"
    client_key '/var/run/ocp-collector/secrets/es-secret/tls.key'
    client_cert '/var/run/ocp-collector/secrets/es-secret/tls.crt'
    ca_file '/var/run/ocp-collector/secrets/es-secret/ca-bundle.crt'
...

Expected results:

Fluentd able to send logs to an Elasticsearch when using passphrase for the private key provided. Then, the configuration generated should be like:

  <match **>
    @type elasticsearch
    @id test
    host elasticsearch
    port 9200
    scheme https
    ssl_version TLSv1_2
    user "#{File.exists?('/var/run/ocp-collector/secrets/es-secret/username') ? open('/var/run/ocp-collector/secrets/es-secret/username','r') do |f|f.read end : ''}"
    password "#{File.exists?('/var/run/ocp-collector/secrets/es-secret/password') ? open('/var/run/ocp-collector/secrets/es-secret/password','r') do |f|f.read end : ''}"
    client_key '/var/run/ocp-collector/secrets/es-secret/tls.key'
    client_cert '/var/run/ocp-collector/secrets/es-secret/tls.crt'
    ca_file '/var/run/ocp-collector/secrets/es-secret/ca-bundle.crt'
    client_key_pass "#{File.exists?('/var/run/ocp-collector/secrets/es-secret/passphrase') ? open('/var/run/ocp-collector/secrets/es-secret/passphrase','r') do |f|f.read end : ''}"
...

The same applies to the {}retry{} section in the fluentd configuration

Workaround

Modify the Elasticsearch Operator for not upgrading automatically and needed manual approval:

$ oc patch subs elasticsearch-operator -n openshift-operators-redhat --type='json' -p='[{"op": "replace", "path": "/spec/installPlanApproval","value": "Manual"}]'
subscription.operators.coreos.com/elasticsearch-operator patched

Verify the Elasticsearch Operator needs manual approval

$ oc get sub elasticsearch-operator -n openshift-operators-redhat-o jsonpath='{.spec.installPlanApproval}' 
Manual

Modify the Cluster Logging Operator for not upgrading automatically and needed manual approval

$ oc patch subs cluster-logging -n openshift-logging --type='json' -p='[{"op": "replace", "path": "/spec/installPlanApproval","value": "Manual"}]'
subscription.operators.coreos.com/cluster-logging patched

Verify the ClusterLogging Operator needs manual approval

$ oc get sub cluster-logging -n openshift-logging -o jsonpath='{.spec.installPlanApproval}'
Manual

Move the Cluster Logging Operator from Managed to Unmanaged

$ oc patch  clusterlogging/instance -n openshift-logging --type='json' -p='[{"op": "replace", "path": "/spec/managementState","value": "Unmanaged"}]'
clusterlogging.logging.openshift.io/instance patched

Edit the collector configmap:

$ oc edit cm collector -n openshift-logging

Add the definition of `client_key_pass`

  <match **>
    @type elasticsearch
    @id elasticsearch    <--- verify the name of this id to be elasticsearch as you have called in the clusterlogfowarder output       
    host elasticsearch
    ...
    client_key '/var/run/ocp-collector/secrets/es-secret/tls.key'
    client_cert '/var/run/ocp-collector/secrets/es-secret/tls.crt'
    ca_file '/var/run/ocp-collector/secrets/es-secret/ca-bundle.crt'
    client_key_pass "#{File.exists?('/var/run/ocp-collector/secrets/es-secret/passphrase') ? open('/var/run/ocp-collector/secrets/es-secret/passphrase','r') do |f|f.read end : ''}"   <---- add this line. Replace <es-secret> by the name giving to the secret used

Do the same for the retry from the buffer file:

  <match retry_elasticsearch>  <--- this is the retry section
    @type elasticsearch
    @id elasticsearch    <--- verify the name of this id to be elasticsearch as you have called in the clusterlogfowarder output
    host elasticsearch
    ... 
    client_key '/var/run/ocp-collector/secrets/es-secret/tls.key'
    client_cert '/var/run/ocp-collector/secrets/es-secret/tls.crt'
    ca_file '/var/run/ocp-collector/secrets/es-secret/ca-bundle.crt'
    client_key_pass "#{File.exists?('/var/run/ocp-collector/secrets/es-secret/passphrase') ? open('/var/run/ocp-collector/secrets/es-secret/passphrase','r') do |f|f.read end : ''}"   <---- add this line. Replace <es-secret> by the name giving to the secret used

Restart the collectors:

$ oc delete pods -l component=collectors -n openshift-logging

Verify that the running fluentd configuration has defined `client_key_pass`:

$ oc -n openshift-logging rsh <collector pod>  grep client_key_pass /etc/fluent/fluent.conf

Verify no more "Enter PEM pass phrase" errors:

$ for pod in $(oc get pods -l component=collector -o name -n openshift-logging); do oc logs $pod -n openshift-logging -c collector |grep -i "Enter PEM";done

Once the bug is fixed, revert to Managed and to Automatic approval of the updates. It should be:

/// Move to automatic the approval of the installPlan
$ oc patch subs elasticsearch-operator -n openshift-operators-redhat --type='json' -p='[{"op": "replace", "path": "/spec/installPlanApproval","value": "Automatic"}]'

/// Move to automatic the approval of the installPlan
$ oc patch subs cluster-logging -n openshift-logging --type='json' -p='[{"op": "replace", "path": "/spec/installPlanApproval","value": "Automatic"}]'

/// Move to Managed the clusterLogging CR
$ oc patch  clusterlogging/instance -n openshift-logging --type='json' -p='[{"op": "replace", "path": "/spec/managementState","value": "Managed"}]'