This tutorial will show you how your instrumentalize your hadoop cluster with MyCloudManager stack to benefit from monitoring, supervision, alerting, log management, backup, etc of your hadoop clusters.
To start, make sure you have launched the stacks in your tenant. If this is not the case, I invite you to watch the following articles :
Start MyCloudManager : http://dev.cloudwatt.com/en/blog/innovation-beta-mycloudmanager-v2.html
Start Hadoop cluster : http://dev.cloudwatt.com/en/blog/deployer-votre-service-hadoop-as-a-service-au-travers-d-un-stack-heat.html
To deploy MyCloudManager’s applications, it is necessary that MyCloudManager can communicate with your instances. You must attach them to the network of your MyCloudManager (r make sure their respective networks can communicate).
Concerning the hadoop cluster deployment, you can choose the network on which it will be deployed. Here’s how:
Clusterspart of your Hadoop console and click
Launch Cluster, choose now the cluster type and version. In this example: plugin
Neutron Management Networkenter here the network of your MyCloudManager then start creating,
curlcommand given to you when adding an instance in MyCloudManager,
After having attached the instances, you can now start the deployement of applications as explained in the MyCloudManager article,
Click on the icons located to the right of each instance to install the corresponding services like Zabbix (for monitoring, supervision, alerting), Ntp, GrayLog (log management)… If the application is not properly installed on the nodes, consider copying the file content ~ /. ssh / authorized_keys2 to ~ / .ssh / authorized_keys and reinstall it.
Instances are now well monitored by MyCloudManager.
Zabbix is configured on the nodes without modifications needs. The metrics of the nodes such as CPU load, utlisation CPU, disk usage … are visible on the interface.
The Log Management (Graylog) is configured to monitor syslogs logs nodes using rsyslog agents. However the Hadoop cluster nodes have other logs to manage separately as DataNodes logs, NameNode logs, Yarn logs, … In order to properly monitor the logs of a Hadoop cluster, we need to add these logs with rsyslog configurations to mount them in Graylog.
wget http://rpms.adiscon.com/v8-stable/rsyslog.repo -O /etc/yum.repos.d/rsyslog.repo yum update rsyslog /etc/init.d/rsyslogd restart
module(load="imfile" mode="inotify") $InputFileName /var/log/hadoop-yarn/hadoop-*.out $InputFileTag yarn: $InputRunFileMonitor
Have Fun and Hack in Peace
Big thanks to Sinh Chung NGUYEN !