Redhat 集群 study notes

gl00ad

backgroud:
OS: centos 5.2
I only want to learn redhat implementation
This is for Load Balancing for static content web application

Question 2: If I want to run RedHat Cluster Suite - cman, rgmanager on Node web20/21/22 to achive High Availability + LVS Load Balance
我想把lvs 和 rhcs 结合起来， 为了说明概念， 不用GFS 和共享磁盘，web 服务器 用本机磁盘。这样可以不用fence 设备。

因为已有lvs做负载平衡，一个节点失败， lvs 会转到另一个节点, High Availability(RHCS)还有必要吗 in this situation？


Question 3:
如果我要用rhcs 的话， web20/21/22 每次只有一个节点active, 唯一同时用的理由是： lvs（load balance) /rhcs(HA)判断节点失败的标准不一样， 我的理解对吗？

Question 1: which is primary/backup Heartbeat Channel?

A) LVS1 eth0 --switch1-- lvs2 eth0
B) lvs1 eth2 -- lvs2 eth2
C) lvs1 eth1 - switch2 -- lvs2 eth1

My understanding is:

A can be primary heartbeat channel, C can be backup heartbeat channel, while B is optional(if need dedicated one), is my understanding right?

[ 本帖最后由 gl00ad 于 2008-10-29 10:08 编辑 ]

kns1024wh

具体的问题加我的QQ交流一下

gl00ad

this is the space for implementation later

Au_Hank

我觉得这个问题的意思是: 那个物理通道可以作为主/备均衡器的心跳通道?
我的理解是B,必须是直接连接的通道.免得因为SWITCH的问题造成主备乱切换.

这里的主/备是指LB的主备,而不是通道的主备. 不知道是否理解正确?

gl00ad

redhat original document:
Periodically, the LVS routers exchange heartbeat messages through the primary external public interface and, in a failover situation, the private interface.

主/备均衡器的心跳通道 have 2 心跳通道, 我觉得 主通道 is A, but I am not sure 备通道. B or C. thanks

Au_Hank

感觉B,C都可以作为BACKUP通道, 但是如果SWITCH2出问题的话,可能引起不必要的切换?

gl00ad

guess you are right, I am just planning what hardware to buy for testing, I will skip switch redundancy, get rid of eth2. thanks

[ 本帖最后由 gl00ad 于 2008-10-28 10:01 编辑 ]

oioilu

看Redhat官方文档，建议使用eth0作为心跳。至于是直连还是通过交换机互联，个人觉得直连比较可靠。但是看坛子上很多人建议使用交换机，原因是怕网线问题。我倒是没这么认为。

gl00ad

Can anyone answer question2, I am really confused by combining LVS and  Redhat Cluster together, especially httpd? thanks

gl00ad

Do you have port 5404 when you run rhel5 cluster cman-openais?

[root@s12 ~]# netstat -an|grep 5405
udp        0      0 192.168.0.12:5405           0.0.0.0:*
udp        0      0 239.192.2.98:5405           0.0.0.0:*
[root@s12 ~]# netstat -an|grep 5404
[root@s12 ~]#

original redhat document

Openais uses three UDP sockets for communications. A multicast receive socket which is bound to the
multicast address. A multicast send socket which is bound to the local IP address but at the port number
specified minus one, and a token socket which is bound to the local IP address. The default port number
is 5405, so a normal system will have two sockets bound to port 5405 and one to 5404.

[ 本帖最后由 gl00ad 于 2008-10-29 20:58 编辑 ]