ORACLE在HP-UX下的系列問題處理(38)
2024-08-29 13:40:02
供稿：網友
OV IT/O: 無法在SG環境中通過防火墻分配模板問題描述
我無法在Service Guard環境中通過防火墻分配模板，同時，安裝代理軟件時沒有任何問題。
為什么會出現這個問題，如何解決？
配置信息
操作系統 - HP-UX
版本 - 10.20
硬件系統 - HP9000
應用程序 - Oracle 7.2.3
解決方法
你最初的問題是由于rpcd向代理程序報告物理結點的源地址。
根據你的防火墻配置的方式，在通過防火墻時，你不能用MC Service Guard配置IT/O。
這個問題可以追溯到這樣的事實，即UDP是傳輸協議，沒有辦法鎖定無連接協議的源地址。
當數據包發送給治理結點時選擇目標地址，操作系統根據治理服務器中的網絡路由表分配源地址。由于同一個網絡號使用相同的主機號，因此使用給定網卡(MC Service Guard)上的第一個地址，我們不能改變這個行為。
這個問題的解決方法是使用傳輸控制協議(TCP)進行傳輸。由于TCP是一個基于連接的協議，因此會話過程中源地址是統一的。假如治理結點在可再定位的地址打開一個會話，那么從治理服務器到治理結點的所有流量（在該會話中）將都以這個可再定位的地址作為源地址。
你也可以考慮重新配置防火墻，接受來自可再定位地址和固定地址的數據包。
當前的技術還無法為你馬上提供一個解決方法。
.........following with all English text ....
PRoblem Description
I cannot distribute templates in a Service Guard environment over the
firewall. At the same time, I don't have any problems installing the agent software.
Why am I having this problem, and do you have a solution?
Configuration Info
Operating System - HP-UX
Version - 10.20
Hardware System - HP9000
application - Oracle 7.2.3
Solution
Your initial problem is due to rpcd reporting back to the agent with the source address of the physical node.
Based upon the way your firewall is configured, you cannot configure IT/O with MC Service Guard while passing through a firewall.
The problem can be traced to the fact that UDP is the transport protocol,and there is no way to lock down the origination address for a connectionless protocol.
When a packet is sent to a managed node, the destination address is selected, and the operating system assigns the source address according to network routing tables in the management server. Since two host numbers exist for the same network number, the first address on a given
Network Interface Card (MC Service Guard) is used. We can't change this behavior.
The solution to this problem is to use Tranmission Control Protocol (TCP) as the transport. Because TCP is a connection-based protocol, the origination address is consistent for the session. If the managed node opens a session to the relocatable address,
all traffic from the
management server to the managed node (in that session) will have the
relocatable address as the source address.
You could also consider reconfiguring the firewalls to allow packets from the relocatable address and the stationary address.
With the current technology, it is not possible to give you an immediate solution.
