Some problems encountered during RedHat + Oracle installation are worth summarizing

Some problems encountered during RedHat + Oracle installation are worth summarizing

Yesterday, a RedHat Linux version 5 + Oracle was installed in the middle of the night. Several things were encountered during the installation. The statement is as follows.

1. you need to configure the kernel parameters before installation. You can refer to the minimum value recommended in the Oracle official documentation to set the parameters. shmmax, kernel. shmall, also referred to many posts on the Internet, the set value is different, at this time, the official document in a sentence worth noting:

"Note: If the current value of any parameter is higher than the value listed in this table, then do not change the value of that parameter ."

That is to say, if the value of an existing parameter is greater than the minimum value recommended in the document, do not change it.

2. including official documents and some practical online posts, all mentioned that you can modify Shell Limits restrictions to improve efficiency. So why?

Generally, we need to add the nproc and nofile values corresponding to the oracle Installation account. There are usually hard and soft values. What exactly do they represent?

In fact, some answers can be obtained from the comments in the/etc/security/limits. conf file:

(1) format <domain> <type> <item> <value>,

<Domain> can be a user name or group name. The wildcard * indicates the default ownership, and the wildcard % indicates the fuzzy match option.

<Type> There are soft and hard values. soft indicates the currently effective setting value of the system, and hard indicates the maximum acceptable setting value of the system.

<Item> there are many items, such as core, stack, nofile, and nproc. nofile indicates the maximum number of files that can be opened, and nproc indicates the maximum number of processes.

3. When you configure the account environment variables, some posts may write a lot. Even if it is PATH or LD_LIBRARY_PATH, different people may load different files during installation, which is indeed quite confusing. But in fact, in my personal opinion, there are more important: ORACLE_HOME, ORACLE_SID, ORACLE_BASE. If you use some sqlplus command line tools, therefore, adding the corresponding tool bin PATH to the PATH is also essential. LD_LIBRARY_PATH, CLASSPATH, and other variables will have some library paths. In addition, NLS_LANG indicates the current language environment, other options may not be required. I listed some of the profile content I added:

4. when you start the Oracle Installation graphical interface, some installation posts often say that the DISPLAY parameter is required, but at least "xhost +" should be used, and xhost is used to control the access permissions of X Server, generally, when hostA is used to log on to hostB to execute the application, hostA is the client for the application, but the graphical interface installed at this time is displayed on hostB. Therefore, you need to run xhost + on hostA first, allow any other user to access the X Server of hostA, or use xhost + ip to access the X Server of the specified ip address. Of course, the above operations are based on the scenario of switching from the root account to the oracle Installation account, for example:

Run xhost + in the root account

Su-oracle

Oracle account execution./runInstaller

If you directly log on to the oracle account for installation, you may not need to perform the preceding operations.

5. I started to use the 11.2.0.4 installation package provided by MOS, but executed. after/runInstaller, no error is prompted, including logs, but the graphic interface is not displayed. After waiting for half a day, I switched back to the original 11.2.0.1 and searched for it. Some of them are similar to my experiences, the reason for this problem is that the installation media is damaged. You may need to download the installation media of 11.2.0.4 again and compare the size to determine it. However, earlier versions can be used, at least indicating that the platform is not a problem, it may be in the installation media. This is a temporary issue.

6. When installing the database, the system will prompt you to configure EM, provided that a LISTENER is required. After I use netca to configure the default LISTENER for LISTENER, the following message is displayed regardless of start, status, and stop:

And Linux Error: 104: Connection reset by peer.

I found that I did not find the cause in many places. Later, according to a post on the Internet, I introduced a MOS article (343295.1), pointing out something similar to "Linux Error: 104: connection reset by peer error often occurs in the new installation process, and three possible causes are proposed:

1. There is possibly an incorrect IP Address specified for the host in the/etc/hosts file.

For example:

192.168.101.101 prod1.us.oracle.com prod1

Whereas the actual ip address for the host "prod1.us.oracle.com" is 192.168.101.110

2. Also, the localhost reference may be incorrect or missing from the/etc/hosts file.

3. Oracle may not have read access to/etc/nsswitch. conf file.

The corresponding methods include:

1. Correct the mapping in the/etc/hosts file by adding/correcting the IP address and/or hostname reference for the Unix/Linux Server (both long and short host names ).

For example:

# Add or edit the/etc/hosts file to include a valid entry for the Server:

192.168.101.110 prod1.us.oracle.com prod1

2. Check that the localhost loopback name and address are correct (both long and short host names ).

An example of a correct localhost entry wocould be as follows:

127.0.0.1 localhost. localdomain localhost

3. Remember to save the/etc/hosts file and then start the listener.

4. Check permissions on/etc/nsswitch. conf file to ensure group and other have read access.

-If necessary change permisions on/etc/nsswitch. conf as follows:

Chmod 644/etc/nsswitch. conf

In my installation, the original listener. ora uses HOST = hostname and then changes it to the actual IP address, which also reports an error. According to the first and second points mentioned above, check/etc/hosts. There is a corresponding ip address and hostname, but this line does not have 127.0.0.1 localhost. It is commented out.

#127.0.0.1 localhost. localdomain localhost4 localhost4.localdomain4

#: 1 localhost. localdomain localhost6 localhost6.localdomain6

172.27.19.56 dcsopenNode1.localdomain dcsopenNode1

So, open the comment and restart the listener. Everything is normal.

As for the third point, no research has been conducted.

Summary:

The above describes some problems encountered during the installation process, including Kernel Parameter Setting Value, limits. conf file, environment variable, and listener startup failure.