ch 10 Monitoring NCNU CSIE 林似真 Stella
NCNU CSIE Stella2010/6/82 ganglia
NCNU CSIE Stella2010/6/83 [ˈgæŋglɪə] KK
NCNU CSIE Stella2010/6/84
NCNU CSIE Stella2010/6/85 outline ● WHY? ● HOW? ● WHAT? ● Ganglia ● Distributed Monitoring System ● Ganglia Architecture – gmond – gmetad – web front-end ● Monitor Intersection ● Ganglia Data Flow ● Distribution and Install ● Example ● Reference ● Q&A
NCNU CSIE Stella2010/6/86 WHY?
NCNU CSIE Stella2010/6/87 Why to Monitor? ● 系統狀態の監控 ● 解決系統の問題 ● 系統或應用程式の調校 ● 系統升級時の依據
NCNU CSIE Stella2010/6/88 HOW?
NCNU CSIE Stella2010/6/89 How to Monitor? ● 單一主機の監控 ● 分散式監控系統
NCNU CSIE Stella2010/6/810 How to Monitor? ● 單一主機の監控 ● commands : top, free, uptime, ps ● Log file ● Shell scripts ● MRTG ●...etc. 分散式監控系統
NCNU CSIE Stella2010/6/811 How to Monitor? ● 分散式監控系統 ● Network utilization ● CPU utilization ● Hardware ● System load ● Disk utilization ●...etc.
NCNU CSIE Stella2010/6/812 Ganglia? ● Nagios ←→ Ganglia ● graph display → easy to see ● show host metrics ● visualize trends
NCNU CSIE Stella2010/6/813 WHAT?
NCNU CSIE Stella2010/6/814 WHAT IS GANGLIA ?
NCNU CSIE Stella2010/6/815 Ganglia - Wiki
NCNU CSIE Stella2010/6/816 Ganglia ● distributed monitoring system ● monitor tool for clusters and gridsclustersgrids ● view live or historical statistics ● CPU load averages ● network utilization ● long-term trends ● graph display
NCNU CSIE Stella2010/6/817 WHAT IS Distributed System ?
NCNU CSIE Stella2010/6/818 Distributed Monitoring System ● Components : 1.daemons that reside on cluster nodes 2.server that collects clusters state information from nodes 3.GUI-based front-end, which provides system activity visulization → gmond → gmetad → web front-end as for Ganglia components :
NCNU CSIE Stella2010/6/819 Ganglia Architecture
NCNU CSIE Stella2010/6/820 gmond ● Ganglia monitoring daemon ● Multithreaded ● Responsibilities : 1.Monitor changes in host states 2.Announce relevant changes 3.Listen to the state of all the other ganglia nodes via a unicast or multicast channel 4.Answer requests for an XML description of the cluster state.
NCNU CSIE Stella2010/6/821 gmetad ● Ganglia meta daemon ● Aggregate data, information ● steps : 1.Poll gmond, clusters, or other gmetad 2.Parse the collected XML 3.Save numeric metrics in RRD(round-robin database) 4.Export aggregated XML over TCP to clients
NCNU CSIE Stella2010/6/822 web front-end ● web interface written in PHP ● Graph generated by gmetad displays historical information ● highly dynamic
NCNU CSIE Stella2010/6/823 monitor interaction ● Gmond runs on each cluster node; gmetad can fail over between nodes.
NCNU CSIE Stella2010/6/824
NCNU CSIE Stella2010/6/825 ● Distribution and Install
NCNU CSIE Stella2010/6/826 example ● official website demo official website demo ● wikimedia demo wikimedia demo
NCNU CSIE Stella2010/6/827 Refernc e
NCNU CSIE Stella2010/6/828 Reference ● Ganglia HowTo Ganglia HowTo ● Ganglia official website Ganglia official website ● wikipedia/Ganglia wikipedia/Ganglia
NCNU CSIE Stella2010/6/829 Q & A
NCNU CSIE Stella2010/6/830 Fin. Now You Can Rest Easy (In Peace ?).