Install Ganglia Monitoring System On Ubuntu 13.10 / Debian 7

The address of original essay: http://www.unixmen.com/install-ganglia-monitoring-system-ubuntu-13-10-debian-7/



Ganglia is a scalable distributed monitoring system for high-performance computing systems such as clusters and Grids. It is based on a hierarchical design targeted at federations of clusters. It leverages widely used technologies such as XML for data representation, XDR for compact, portable data transport, and RRDtool for data storage and visualization.

It uses carefully engineered data structures and algorithms to achieve very low per-node overheads and high concurrency. The implementation is robust, has been ported to an extensive set of operating systems and processor architectures, and is currently in use on thousands of clusters around the world. It has been used to link clusters across university campuses and around the world and can scale to handle clusters with 2000 nodes.

Be mindful that Ganglia will only help you to view the performance of your servers, and it doesn’t tweak or improve the performance. In this tutorial, we are going to implement Ganglia Monitoring Tool on Ubuntu 13.10 server and let us use Ubuntu 13.04 as our Monitoring target. Though it was tested on Ubuntu 13.10, the same method should work on Debian 7 and other Ubuntu versions as well.

Install Ganglia On Ubuntu 13.10

Before proceeding to install Ganglia, you have to complete the following tasks.

Make sure your Server has a properly installed and configured LAMP stack. To install and configure LAMP server, refer the following link.

Install LAMP Server On Ubuntu 13.10

If you’re using Debian, refer the following link.

Install LAMP Server On Debian 7

Ganglia consists of two main daemons called gmond (Ganglia Monitoring Daemon) and gmetad (Ganglia Meta Daemon), a PHP-based web front-end and a few other small utilities.

Ganglia Monitoring Daemon (gmond):

Gmond runs on each node you want to monitor and monitor changes in the host state, announce relevant changes, listen to the state of all other ganglia nodes via a unicast or multicast channel and answer requests for an XML description of the cluster state.

Ganglia Meta Daemon (gmetad):

Gmetad runs on the master node which gathers all information from the client nodes.

Ganglia PHP Web Front-end:

It displays all the gathered information from the clients in a meaningful way like graphs via web pages.

Ganglia Installation On Master node

Install Ganglia using command:

$ sudo apt-get install ganglia-monitor rrdtool gmetad ganglia-webfrontend

During installation, you’ll be asked to restart apache service to activate the new configuration. Click Yes to continue.

sk@server: ~_001

Configure Master node

Now copy ganglia configuration file /etc/ganglia-webfrontend/apache.conf to /etc/apache2/sites-enabled/directory as shown below.

$ sudo cp /etc/ganglia-webfrontend/apache.conf /etc/apache2/sites-enabled/ganglia.conf

Then edit file /etc/ganglia/gmetad.conf,

$ sudo nano /etc/ganglia/gmetad.conf

Find the following line and modify as shown below.

data_source "my cluster" 50 192.168.1.101:8649

As per the above line, the logs will be collected from each node every 50 seconds. Also, you can assign a name for your client groups. In my case, I use the default group name “my cluster”. Here 192.168.1.101 is my master node IP address.

Save and close the file.

Edit file /etc/ganglia/gmond.conf,

$ sudo nano /etc/ganglia/gmond.conf

Find the following sections and modify them with your values.

[...]
cluster {
  name = "my cluster"  ## Name assigned to the client groups
  owner = "unspecified"
  latlong = "unspecified"
  url = "unspecified"
}

[...]

udp_send_channel   {
#mcast_join = 239.2.11.71 ## Comment
  host = 192.168.1.101   ## Master node IP address
  port = 8649
  ttl = 1
}

[...]

udp_recv_channel {
  port = 8649
}

/* You can specify as many tcp_accept_channels as you like to share
   an xml description of the state of the cluster */
tcp_accept_channel {
  port = 8649
}

[...]

The changes in the above configuration file show that the master node which has IP address 192.168.1.101 will collect data from all nodes on tcp and udp port 8649.

Save and close the file. Then start ganglia-monitor, gmetad and apache services.

$ sudo /etc/init.d/ganglia-monitor start
$ sudo /etc/init.d/gmetad start
$ sudo /etc/init.d/apache2 restart

Ganglia Installation On Clients

Install the following package for each client you want to monitor.

On Debian / Ubuntu clients:

$ sudo apt-get install ganglia-monitor

On RHEL based clients:

# yum install ganglia-gmond

Configure Clients

Edit file /etc/ganglia/gmond.conf,

$ sudo nano /etc/ganglia/gmond.conf

Make the changes as shown below.

[...]

cluster {
  name = "my cluster"     ## Cluster name
  owner = "unspecified"
  latlong = "unspecified"
  url = "unspecified"

[...]

udp_send_channel {
  #mcast_join = 239.2.11.71   ## Comment
  host = 192.168.1.101   ## IP address of master node
  port = 8649
  ttl = 1
}
## Comment the whole section
/* You can specify as many udp_recv_channels as you like as well.
udp_recv_channel {
  mcast_join = 239.2.11.71
  port = 8649
  bind = 239.2.11.71
}
*/

tcp_accept_channel {
  port = 8649
}

[...]

Save and close the file. Next, restart ganglia-monitor service.

On Debian based systems:

$ sudo /etc/init.d/ganglia-monitor restart

On RHEL based systems:

# service gmond restart

Access Ganglia web frontend

Now point your web browser with URL http://ip-address/ganglia. You should see the client node graphs.

Ganglia:: unspecified Cluster Report - Mozilla Firefox_002

To view a particular node graphs, select the particular node you want from the Grid Choose Node drop-down box.

For example, i want to see the graphs of Ubuntu client which has IP address 192.168.1.100.

Ganglia:: unspecified Cluster Report - Mozilla Firefox_005

Graphs of my Ubuntu client (192.168.1.100) client:

Ganglia:: 192.168.1.100 Host Report - Mozilla Firefox_004

Client Node View:

Ganglia:: 192.168.1.100 Node View - Mozilla Firefox_006

Server Node view:

Ganglia:: 192.168.1.101 Node View - Mozilla Firefox_007

As you see in the above outputs, my client node (192.168.1.101) is down and server node (192.168.1.100) is up.

Source & Reference Links:

Ganglia Homepage


  • 0
    点赞
  • 0
    收藏
    觉得还不错? 一键收藏
  • 0
    评论
1 目标检测的定义 目标检测(Object Detection)的任务是找出图像中所有感兴趣的目标(物体),确定它们的类别和位置,是计算机视觉领域的核心问题之一。由于各类物体有不同的外观、形状和姿态,加上成像时光照、遮挡等因素的干扰,目标检测一直是计算机视觉领域最具有挑战性的问题。 目标检测任务可分为两个关键的子任务,目标定位和目标分类。首先检测图像中目标的位置(目标定位),然后给出每个目标的具体类别(目标分类)。输出结果是一个边界框(称为Bounding-box,一般形式为(x1,y1,x2,y2),表示框的左上角坐标和右下角坐标),一个置信度分数(Confidence Score),表示边界框中是否包含检测对象的概率和各个类别的概率(首先得到类别概率,经过Softmax可得到类别标签)。 1.1 Two stage方法 目前主流的基于深度学习的目标检测算法主要分为两类:Two stage和One stage。Two stage方法将目标检测过程分为两个阶段。第一个阶段是 Region Proposal 生成阶段,主要用于生成潜在的目标候选框(Bounding-box proposals)。这个阶段通常使用卷积神经网络(CNN)从输入图像中提取特征,然后通过一些技巧(如选择性搜索)来生成候选框。第二个阶段是分类和位置精修阶段,将第一个阶段生成的候选框输入到另一个 CNN 中进行分类,并根据分类结果对候选框的位置进行微调。Two stage 方法的优点是准确度较高,缺点是速度相对较慢。 常见Tow stage目标检测算法有:R-CNN系列、SPPNet等。 1.2 One stage方法 One stage方法直接利用模型提取特征值,并利用这些特征值进行目标的分类和定位,不需要生成Region Proposal。这种方法的优点是速度快,因为省略了Region Proposal生成的过程。One stage方法的缺点是准确度相对较低,因为它没有对潜在的目标进行预先筛选。 常见的One stage目标检测算法有:YOLO系列、SSD系列和RetinaNet等。 2 常见名词解释 2.1 NMS(Non-Maximum Suppression) 目标检测模型一般会给出目标的多个预测边界框,对成百上千的预测边界框都进行调整肯定是不可行的,需要对这些结果先进行一个大体的挑选。NMS称为非极大值抑制,作用是从众多预测边界框中挑选出最具代表性的结果,这样可以加快算法效率,其主要流程如下: 设定一个置信度分数阈值,将置信度分数小于阈值的直接过滤掉 将剩下框的置信度分数从大到小排序,选中值最大的框 遍历其余的框,如果和当前框的重叠面积(IOU)大于设定的阈值(一般为0.7),就将框删除(超过设定阈值,认为两个框的里面的物体属于同一个类别) 从未处理的框中继续选一个置信度分数最大的,重复上述过程,直至所有框处理完毕 2.2 IoU(Intersection over Union) 定义了两个边界框的重叠度,当预测边界框和真实边界框差异很小时,或重叠度很大时,表示模型产生的预测边界框很准确。边界框A、B的IOU计算公式为: 2.3 mAP(mean Average Precision) mAP即均值平均精度,是评估目标检测模型效果的最重要指标,这个值介于0到1之间,且越大越好。mAP是AP(Average Precision)的平均值,那么首先需要了解AP的概念。想要了解AP的概念,还要首先了解目标检测中Precision和Recall的概念。 首先我们设置置信度阈值(Confidence Threshold)和IoU阈值(一般设置为0.5,也会衡量0.75以及0.9的mAP值): 当一个预测边界框被认为是True Positive(TP)时,需要同时满足下面三个条件: Confidence Score > Confidence Threshold 预测类别匹配真实值(Ground truth)的类别 预测边界框的IoU大于设定的IoU阈值 不满足条件2或条件3,则认为是False Positive(FP)。当对应同一个真值有多个预测结果时,只有最高置信度分数的预测结果被认为是True Positive,其余被认为是False Positive。 Precision和Recall的概念如下图所示: Precision表示TP与预测边界框数量的比值 Recall表示TP与真实边界框数量的比值 改变不同的置信度阈值,可以获得多组Precision和Recall,Recall放X轴,Precision放Y轴,可以画出一个Precision-Recall曲线,简称P-R
评论
添加红包

请填写红包祝福语或标题

红包个数最小为10个

红包金额最低5元

当前余额3.43前往充值 >
需支付:10.00
成就一亿技术人!
领取后你会自动成为博主和红包主的粉丝 规则
hope_wisdom
发出的红包
实付
使用余额支付
点击重新获取
扫码支付
钱包余额 0

抵扣说明:

1.余额是钱包充值的虚拟货币,按照1:1的比例进行支付金额的抵扣。
2.余额无法直接购买下载,可以购买VIP、付费专栏及课程。

余额充值