Chinaunix首页 | 论坛 | 博客
  • 博客访问: 1895776
  • 博文数量: 389
  • 博客积分: 7877
  • 博客等级: 少将
  • 技术积分: 4521
  • 用 户 组: 普通用户
  • 注册时间: 2007-12-10 14:02
文章分类

全部博文(389)

文章存档

2024年(1)

2022年(1)

2021年(1)

2020年(1)

2019年(1)

2018年(3)

2017年(6)

2016年(4)

2015年(8)

2014年(15)

2013年(31)

2012年(19)

2011年(47)

2010年(33)

2009年(105)

2008年(109)

2007年(4)

分类:

2008-11-14 16:14:53

1. Add cluster
2. Install the grumpy application(/usr/local/grumpy)
3. Create share ecvg(enhanced concurrent VG) GRUMPYvg
4. Add a service IP label to /etc/hosts file on both nodes, called grumpserv
5. Define grumpserv to HACMP as a service IP address.
6. Add a Resource Group to HACMP called grumpyRG.
7. Add the service IP address and VG to the Resource Group.
8. Add a Resource Group to HACMP called webRG, then create a resource group parent/child relationship,
   where grumpy is the child and webRG is the parent.

add application server to RG on both

9. Sync the cluster, share ecvg will show on the highest priority node.
10. Use lsvg to check the state of the volume group GRUMPYvg, (lsvg -o)
11. Create a jfs2 filesystem called /grumpyfs on GRUMPYvg.(100MB is enough)
12. Check that /grumpyfs has been mounted on the highest priority node.

13. Define an application server for grumpy called grumpyAPP. Include grumpystart and grumpystop as the application
    stop/start scripts.
14. Add the application server to the resource group(grumpyRG)
15. Add application monitor(start-up/long-running/cunstom/process)
16. Sync the cluster, check grumpy availability(if all node of cluster work properly with out reboot itself
    and there are log files at /grumpyfs. This result is right!)


得用到parent/child dependencies, 作为父资源的webRG也需要takeover到另外一个node上。这里可能得考虑使用node location
dependencies.

定义application monitor时,需要定义两组,grumpyRG和webRG都需要takeover到另外一个node上。

* one cluster: testing_cl
 
* two resource group: grumpyRG, webRG

* two node belong to cluster testing_cl: node1, node2

* one share ecvg: GRUMPYvg

* grumpyRG contains service IP(grumpserv), share ecvg(GRUMPYvg), and application server(grumpyAPP)

* Most important things: Application Monitor(when the highest priority node have problem, the monitor can find out
  that, and make all the RG takeover to the lowest priority node, because grumpy start need httpd daemon, so when
  application takeover to the lowest priority node, it should first startup httpd daemon and then grumpy will startup
  as we want)
阅读(1514) | 评论(0) | 转发(0) |
给主人留下些什么吧!~~