Description of problem: We need to check if the current metrics role supportsconfiguring the hosts to ship the data to elasticsearch instance that is not based on OpenShift. Steps to Reproduce: 1. Setting up elasticsearch instance 2. Creating project.ovirt-metrics and project.ovirt-logs dynamic index templates. 3. Log in to the Manager machine using SSH. 4. Copy config.yml.example to create config.yml: # cp /etc/ovirt-engine-metrics/config.yml.example /etc/ovirt-engine-metrics/config.yml.d/config.yml 5. Edit the ovirt_env_name and elasticsearch_host parameters in config.yml and add the following variables: use_omelasticsearch_cert: false viaq_metrics_store: false 6. Save the file. These parameters are mandatory and are documented in the file. 7. Deploy Collectd and Rsyslog on the hosts: # /usr/share/ovirt-engine-metrics/setup/ansible/configure_ovirt_machines_for_metrics.sh 8. Import the pre-defined dashboards to Kibana(User will be prompt to choose the index to use for the dashboards, and he should choose the project.ovirt-metrics. and project.ovirt-logs indexes from the drop down list. Expected results: Data should be saved to elasticsearch and visible in Kibana. Additional info:
Additional variables that needs to be set in the config.yml file: Set if no certificates required: use_omelasticsearch_cert: false If not using https add the also following: rsyslog_elasticsearch_usehttps_metrics: off rsyslog_elasticsearch_usehttps_logs: off If certificate are required user will need to specify their location, Default locations: # Where to find the CA certificate used to communicate with Elasticsearch rsyslog_elasticsearch_ca_cert_path: '/etc/rsyslog.d/elasticsearch_ca_cert.pem' # Where to find the client certificate used to communicate with Elasticsearch rsyslog_elasticsearch_client_cert_path: '/etc/rsyslog.d/elasticsearch_client_cert.pem' # Where to find the client certificate used to communicate with Elasticsearch rsyslog_elasticsearch_client_key_path: '/etc/rsyslog.d/elasticsearch_client_key.pem'
*** Bug 1693744 has been marked as a duplicate of this bug. ***
Update to step 2 in comment #1 The dynamic index templates names should be: project.ovirt-metrics-<ovirt_env_name> project.ovirt-logs-<ovirt_env_name>
Yes, we do need to support customers who have existing ELK stacks for logging and metrics. We'll come up with a tighter set of requirements for what we want to support, and a reasonable set to test against to limit the infinite field. Let's re-open this, and track as an RFE to deliver this feature.
I was able to test it following the steps: 1. Installed ElasticSearch 5.6.16 in a fresh vm following: https://www.elastic.co/guide/en/elasticsearch/reference/5.6/_installation.html 2. On ES machine append the following "network.host: 0.0.0.0" on /etc/elasticsearch/elasticsearch.yml # cat "network.host: 0.0.0.0" >> /etc/elasticsearch/elasticsearch.yml 3. Followed this bz description and comment #1 for metrics side 4. After deploying rsyslog and collectd, I manually edit /etc/rsyslog.d/30-elasticsearch.conf following varialbes: # usehttps="off" (it was "on") # allowUnsignedCerts="off" (deleted this line) Results: Data was being pushed to ES instance outside ocp env
Hi Guilherme, sharing a draft of the KB article with you, please let me know if this provides the procedure that is needed: https://docs.google.com/document/d/1pEZCIVud4a8F_qiJb17qWvQqzH-rVkWZiMYOeLdNxOo/edit?usp=sharing
It looks good to me Eli! It does provide the steps needed.
I just want to complement my comment #9 about the installation of Elastic Search. The tar installation method, presented in the tutorial link, may have some issues depending of the environment. A better approach is to use either the repositories or rpm here: https://www.elastic.co/guide/en/elasticsearch/reference/5.6/rpm.html#rpm-repo Also there is a typo on step 2, the cmd should be: # cat "network.host: 0.0.0.0" >> /etc/elasticsearch/elasticsearch.yml
(In reply to Guilherme Santos from comment #14) > I just want to complement my comment #9 about the installation of Elastic > Search. > The tar installation method, presented in the tutorial link, may have some > issues depending of the environment. A better approach is to use either the > repositories or rpm here: > https://www.elastic.co/guide/en/elasticsearch/reference/5.6/rpm.html#rpm-repo > > Also there is a typo on step 2, the cmd should be: > # cat "network.host: 0.0.0.0" >> /etc/elasticsearch/elasticsearch.yml Typo again: # echo "network.host: 0.0.0.0" >> /etc/elasticsearch/elasticsearch.yml