des plateformes hadoop bertrand dechoux - …...cédric carbone - @carbone bertrand dechoux -...
TRANSCRIPT
Supervision et administrationdes plateformes HadoopOù en est-on ? Comment relier cela à la supervision du SI ?
Cédric Carbone - @carboneBertrand Dechoux - @BertrandDechoux
Hadoop
“perdre une machine est un non-événement”
mais
avez vous besoins de vos applications ?
Hadoop 2 : stockage haute dispo
NameNode NameNode
DataNode DataNode DataNode DataNode
DataNode DataNode DataNode
Hadoop 2 : calcul haute dispo
RessourceManager
RessourceManager
NodeManager
NodeManager
NodeManager
NodeManager
NodeManager
NodeManager
NodeManager
Mesurer, pourquoi ?
● Notifier les erreurs, matérielles ou applicatives
● Diagnostiquer les erreurs, matérielles ou applicatives
● Planifier les ressources nécessaires
Apache Ambari, pour Hadoop
● Provision a Hadoop Clustera. Ambari provides an easy-to-use, step-by-step wizard for installing Hadoop services across any number of hosts.b. Ambari handles configuration of Hadoop services for the cluster.
● Manage a Hadoop Clustera. Ambari provides central management for starting, stopping, and reconfiguring Hadoop services across the entire cluster.
● Monitor a Hadoop Clustera. Ambari provides a dashboard for monitoring health and status of the Hadoop cluster.b. Ambari will send emails when your attention is needed (e.g., a node goes down, remaining disk space is low, etc).
Ambari consists of a set of RESTful APIs and a browser-based management interface.
Cloudera Manager, pour Hadoop
Apache Ambari / Cloudera Manager
● Both Apache Ambari and Cloudera Manager do the job of managing and monitoring your cluster
● Difference on the Security layer (Sentry vs Apache Ranger) and on the Business Model (OSS vs Proprietary)
● Cloudera Manager started 3 years before Ambari but the feature-gap is quickly closing.
● However, with both Ambari / Cloudera Manager you cannot dig into your application => you need to add a dedicated monitoring/log management tool
Log centralisation
Logmatic
Monitoring
● série temporelle
● RRD○ “Round Robin Database”
Monitoring : build or buy
hardware
OS
applications
build
buy
Cédric Carbone - @carboneBertrand Dechoux - @BertrandDechoux