ITPub博客

首页 > 数据库 > Oracle > 安装Oracle Enterprise Manager 10g Grid Control Release 10.1.0.3

安装Oracle Enterprise Manager 10g Grid Control Release 10.1.0.3

原创 Oracle 作者:zhulch 时间:2007-09-18 09:16:48 0 删除 编辑
..................[@more@]

安装了一个Oracle Enterprise Manager 10g Grid Control Release 10.1.0.3 , 并在一个测试DB上安装上了agent .

没感觉好玩呢..据说这个option 还要单独收费,真不值!

以下是这个软件的readme.

__________________________________________

Oracle Enterprise Manager 10g
Grid Control Release 10.1.0.3
Readme
Last updated: July 12, 2004



Table of Contents

1. General
2. Oracle Management Service/Repository/Management Agent
3. Monitoring System
4. Blackouts
5. Jobs
6. Database Management
7. Configuration Management Pack
8. Application Server Management
9. Application Service Level Management
10. Oracle Collaboration Suite
11. EM2Go
12.
Enterprise Manager 9i to 10.1 Migration


1. General

1.1 Introduction
This Readme identifies differences between the delivered Oracle Enterprise Manager 10g Grid Control Release 10.1.0.3. product and its documented functionality.

1.2 Installation

1.2 Installation
1.2.1 Resuming Agent Installations After Failure
While performing "Complete Enterprise Manager with a Database", "Complete Enterprise Manager in an Existing Database" or "Additional Management Service" install, if the agent install fails, you can resume the Management Agent install after exiting Oracle Universal Installer by performing the following steps :

a) Set the environment variable EM_REPOS_PWD to the value of sysman schema password entered in the OMS install, using "setenv" for C-Shell and export for Korn shell .
b) If the install performed was "Complete Enterprise Manager with a Database" or "Complete Enterprise Manager in an Existing Database", set the environment variable EM_REPOS_DBSYSPWD to the value of sys schema password
c) Execute the installer using the following command:

runInstaller -responseFile /agent/agent.rsp

Alternately, you can run the installer in GUI (interactive) mode and invoke the "Additional Management Agent" installation for the agent home. The agent home is typically created at the same level as the Enterprise Manager home. However, it may be nested within the Enterprise Manager home in cases where the parent directory of the Enterprise Manager home cannot be written onto.


1.2.2 Installation of the OMS and Agent on machines with multiple network interfaces and/or virtual hosts
Machines with multiple network interfaces or virtual hosts bound to single or multiple network adapters are accessible by more than one IP address. The Enterprise Manager installation for both the Management Service and the Management Agent will always select the primary machine alias.

1.2.2.1 Changing Management Service IP Address
Changing Management Service IP address after initial installation is not currently supported without deinstalling and reinstalling the Management Service.


1.2.2.2 Changing the Management Agent IP address after initial installation
If the existing host name remains unchanged but resolves to a new IP address, then no changes are required.

If the Agent host is assigned a new machine name (a new primary alias that resolves to the current IP address, or a new IP address), then the following steps will rename the Agent on the managed node.

a) Create a backup of the existing targets.xml
b) Edit the existing targets.xml, replacing the old host name with the new host name
c) Alter the Management Agent configuration file:

Edit the EMD_URL property in
$ORACLE_HOME/sysman/config/emd.properties to reflect the new host name.

d) Shut down and restart the Management Agent
e) Verify that the new host name appears in the Enterprise Manager console along with its associated targets.



Note:
i)The targets associated with the old host name will continue to be displayed in the Enterprise Manager console. If the historical data are not required, it is recommended that the targets associated with the old host name be removed using the Enterprise Manager console UI.
ii) In a RAC environment the emd.properties file would be under $ORACLE_HOME//sysman/config directory.

1.2.3 Retrying Secure Agent action
During the configuration portion of the agent install, the Secure Agent action may fail if the OMS is down or busy processing other requests.

This operation can be retried multiple times from the installer by retrying the "Secure Agent action" until the agent is secured successfully. Another alternative is to complete the installation with this failure and use:


emctl secure agent

at the command line to secure the agent.

1.2.4 Securing OMS via EMCTL Changes Upload Port
When using non-default ports, the command

emctl secure oms

does not retain the console upload ports specified at install time. The ports are always defaulted to 4889 (unsecure) and 4888 (secure). This is bug 3346036

To workaround this problem, set:

EM_UPLOAD_PORT to oracle.sysman.emSDK.svlt.ConsoleServerPort in $ORACLE_HOME/sysman/config/emoms.properties.

EM_UPLOAD_HTTPS_PORT in $ORACLE_HOME/sysman/config/emoms.properties should be set to the port number being used for secure upload.

After setting these variables, rerun emctl secure oms.


Note that in a RAC environment the emoms.properties file would be under $ORACLE_HOME//sysman/config directory.

1.2.5 De-installation
If you de-install the Enterprise manager 10g ORACLE_HOME using Oracle universal installer (OUI), it de-registers the ORACLE_HOME from the central inventory and the oratab file. It may, however, leave some files within the ORACLE_HOME. If the ORACLE_HOME has been successfully removed (it can be verified from OUI by clicking on 'Installed products' button) one can manually delete the files using the appropriate command on the Operating System.

The proper sequence of de-installation is:
a) Shutdown the repository database (if it is a part of the de-installation)
b) Shutdown the database listener
c) Shutdown Apache using "opmnctl stopall"
d) Shutdown Application Server Console using "emctl stop em"
e) Shutdown agent in the chained agent home using "emctl stop agent".
f) Run the Oracle Universal Installer and select the ORACLE_HOME to de-install. It would then be removed from the Central Inventory.
g) Remove stray files from the ORACLE_HOME by appropriate operating system command. .


1.2.6 Problems With Agent Shutdown Operation on NT 4.0

If you are running the agent on a Windows NT 4.0 host, you may find that the agent does not shutdown or recover gracefully from a timeout. This may lead to an inconsistency between the status reported by ‘emctl’ and the status reported by Windows service control manager. To workaround this issue the following internal command “emctl istop agent” can be issued to shutdown the agent. You may then bring up the agent using your Windows service control manager (Bug 3320183)

1.2.7 Verify NLS settings are configured properly
If you launch opmnctl or emctl in non English environment, make sure the OS default locale and the NLS_LANG settings are configured properly. (Bug 3128644)

1.2.7.1 OS default locale

a) Make sure LC_ALL or LANG environment variables are set with the appropriate value. To check the current setting, issue locale command:
% locale for the specific value in each OS, refer to the platform specific documents.



1.2.7.2 Make sure NLS_LANG environment variable is set with the appropriate and compatible value with the OS default locale setting and the repository database character set. For the specific value for the language or the character set, refer to the Globalization Support Guide of the Oracle product you are using.
If the platform is Microsoft Windows based Operating System, the default NLS_LANG setting in the registry should be used as it is. You usually do not have to change the value.



1.2.8 National Language Support: Performance Analysis Findings Return As Garbled Text
If the NLS settings of the Enterprise Manager repository database are different from the NLS settings of a managed database target, the Performance Analysis findings may return as garbled text, e.g. question marks. The workaround is to click on the finding text, which will display the finding in the language of the database target (Bug 3351589)



1.2.9 Changing the agent port for a secured agent
You can change the listen port of an agent by editing the EMD_URL property in the $OH/sysman/config/emd.properties file. If you do this for an agent that has already been secured, you will need to resecure the agent (via emctl secure agent). If you do not resecure the agent, agent uploads to the OMS will fail - data will back up at the agent, and the OMS will not be able to contact the agent to retrieve real time metric values (Bug 2814831). In a RAC environment the emd.properties file would be under $ORACLE_HOME//sysman/config directory.





1.2.10 OMS Fails to Secure

If all registration passwords are removed, the OMS will not secure any new agents. You will get the following error:



The OMS is not set up for Enterprise Manager Security.



To workaround this issue, login to Enterprise Manager as a Super Administrator and add a new registration password from the Setup -> Registration Passwords page. Once this is finished, resecure the agent using the emctl secure agent command. (Bug 3356147)



1.2.11 Use of a RAC database as 10g Grid Repository
The database provided in the installation of the 10g Grid Management Server is a single instance 9.0.1.5 database. In order to have a real application cluster instance, the installation of the 10g Grid Management Server must be done using the using an existing database option.

In the course of the installation, the installer will request a SID to be specified. Chose one instance of the Oracle RAC and specify its SID. When prompted for the location of the 10g Grid management tablespaces, insure that the location is one mutually accessible by all members of the RAC.

Upon completion of the installation, edit the $ORACLE_HOME/sysman/config/emoms.properties file. Alter the property of the name oracle.sysman.eml.mntr.emdRepConnectDescriptor to represent the SERVICE_NAME of the RAC instead of the SID that will be found there.

Restart the OMS.

Subsequent additions of other Management Servers that wish to use the RAC database should perform the installation choice of an 'Additional Management Service'. When prompted for the connection information, they too should specify the SID of one of the database instances comprising the RAC. After the installation has completed, the emoms.properties file in that newly added 10g Grid Management Server's home should be altered in the same manner as in the initial installation. (Bug 3259014)



1.2.12 Oracle Universal Installer Does Not Start Up on Windows 2000

Oracle Universal Installer may not start up properly (appear and then disappear) on Windows 2000 with Service pack 4. This is
due to an underlying JRE bug. You need to run 'installwin32setup.exe -J-Dsun.java2d.noddraw=true' to workaround the issue.



1.2.13 Bug 3417663: This problem occurs when an Oracle Database 10g with Oracle Universal Installer 10.1.0.2.0 or earlier is cloned to a target location on the same host as the source home. Deinstalling the cloned home using OUI will remove some files from the source home and may cause it to be in a non-functional state. This applies to cloning performed by both Oracle Grid Control and Oracle DB Control. The bug has been fixed in Oracle Universal Installer Version 10.1.0.3.0 and will not affect products certified and installed with that version.



1.2.14 Bug 3609193: In non English language environment except Japanese (Brazilian Portuguese, French, German, Italian, Korean, Simplified Chinese, and Spanish), Grid Control online help prints out-of-dated translated contents of "Oracle9iAS Container for Java".(bug#3609193)


In order to work around this issue, the user replaces the out-of-dated translated version of oc_4j_help_.jar files with the English one (oc_4j_help.jar). Specific steps would include:



For UNIX:

As user oracle or the appropriate user id:

% cd $ORACLE_HOME/j2ee/OC4J_EM/applications/em/em/online_help/oc_4j

Perform the following copy command for each .

cp -p oc_4j_help.jar oc_4j_help_.jar

The following b-shell script will do this automatically:

for file in oc_4j_help_*.jar; do

cp -p oc_4j_help.jar $file

done

For Windows:

Perform the same copy operations with Windows Explorer.

1.2.15 File Size Limit on FAT File System (Windows)
On Windows, if the repository database is located on a FAT file system, there is an OS-mandated file limit of 4GB. Even if datafiles are set to auto-extend, they will not be able to exceed 4GB because of the OS limitation. The workaround is to add a new datafile once the 4GB limit is reached. This is not an issue when the repository database is located on NTFS file systems. (Bug 3748521)
1.2.16 Cannot Disable Management Packs for RAC Database and ASM Targets
Disabling access to individual packs using the Management Pack Access interface does not work for cluster database and ASM targets. (Bug 3751881) Note that some cluster database and ASM pages may get blocked because they have the same links as the single-instance equivalent pages. However, pages that are cluster database or ASM specific do not get blocked.

2. Oracle Management Service / Repository / Oracle Management Agent

2.1 OMS and Repository

2.1.1 Securing Multiple Oracle Management Services Should Not Be Done Simultaneously
The first time you secure any Oracle Management Service, you should not attempt to secure more than one Oracle Management Service at a time. After one Oracle Management Service has been successfully secured, you can secure multiple Oracle Management Services at one time. (Bug 3335221)

2.1.2 Queries to Some Views May Hang

Queries to the following views may hang due to RDBMS bug 3243901.

mgmt$metric_details

mgmt$metric_daily
mgmt$metric_hourly
em$rt_loopback_details_1day
em$rt_metrics_details_1day
em$rt_metrics_details_7day
em$rt_metrics_details_31day
em$rt_loopback_details_7day
em$rt_loopback_details_31day

The shadow process corresponding to this database session will consume a lot of memory and CPU. The fix to this problem is to kill that database session and recompile these views. Alternatively use a 9204 database for your management repository.

2.1.3 Out of Memory Errors Under Heavy Load

Under heavy load, especially when connecting to multiple target databases, the Oracle Management Service may encounter out-of-memory errors and restart. When an Oracle Management Service restarts, there is no loss of state in the repository, but users logged in to that Management Service will be asked to login again. If the restarts are frequent, do the following for each installed Management Service:

a) Stop the oms;

$ORACLE_HOME/opmn/bin/opmnctl stopall

(Note: it is important to use opmnctl stopall here rather than the usual "emctl stop oms" since we need to
completely stop the associated Application Server instance as well)

b) Edit $ORACLE_HOME/opmn/conf/opmn.xml
Search for the OC4J_EM section, which is a line that looks like this:

Edit the "java-options" data tag under this process type. Before edit, this line should look similar to this:


Change the -Xmx option from 512M to 1024M, and the -XX:MaxPermSize option from 64M to 256M. Also add the
-Xnoclassgc option.

After edits, the line should look like this:

c) Restart the Oracle Management Service using emctl start oms. (Bug 3272105)

2.1.4 Oracle Management Service Deadlocks

During intense OMS activity, two or more threads in the OMS may deadlock. As a result, the Oracle Management Service interface pages may become unresponsive, or the OMS may not upload data from agents. This is a Oracle jdbc issue, and a patch for this is forthcoming. Please check MetaLink for patch availability. (Bug 3347676)

2.1.5 Security Considerations When Adding Granting "Manage Group" Privileges
As an Enterprise Manager super administrator, use caution when assigning group privileges to other Enterprise Manager administrators. For example, if you grant a user "Full" privileges for a particular group, note that the user who created the group could add additional targets to the group and indirectly give other users full privileges to these targets. In other words, any users that have full privileges on the group will also have full privileges on targets added to the group by the group creator.

In general, Enterprise Manager super administrators should make sure that no users have higher privileges on a group than the user or users who manage the group. (Bug 3217302)

 
2.1.6 Identifying the repository database in the 'Management Services and Repository Overview' page
In the 'Management System' tab, the 'Management Services and Repository Overview' page shows information about the repository database used by Enterprise Manager. In this page, the 'Database' field typically displays the name of the repository database. However, in some cases you may see a connect descriptor instead of the database name, even though the database itself is a managed target within Enterprise Manager. To identify the repository database, Enterprise Manager queries the gv$instance view of the repository database, then tries to find a managed database target whose values for HOST, PORT, and SID exactly match (including case) the values in gv$instance. In cases wherein no match is found, Enterprise Manager displays the connect descriptor for the repository database. To fix the problem, do the following:
a) Query the gv$instance view of the repository database and find out the values for HOST, PORT, and SID
b) In Enterprise Manager, go to the home page of the database target that has the repository. Click on the 'Monitoring Configuration' link and verify that the HOST, PORT, and SID values exactly match (including case) the values from the gv$instance view.

2.2 Management Agent

2.2.1 Troubleshooting Agent Upload Problems
If data from the agent is not arriving at the OMS/console, the agent trace file might provide a hint about the problem.

In the Agent's oracle home directory, look for the sysman/log/emagent.trc file. In this file, look for messages with the string "ERROR upload:" in them. There is one specific case which points to an OMS that is overloaded and is taking longer time than expected to load the files. The entries in the trace file should look like this:

2003-10-31 15:48:45 Thread-1499 ERROR upload: FxferSend: HTTP Error=-2 reading request response from:

http://ap401ses.us.oracle.com:4889/em/upload/

2003-10-31 15:48:45 Thread-1499 ERROR upload: Failed to upload file A0000002.xml , ret = -2

If this happens, you may be able to improve the situation somewhat by increasing the timeout of the Apache server in the OMS home.

In the oracle home of the OMS, edit the $OH/Apache/Apache/conf/httpd.conf file and change the Timeout setting from

Timeout 300
to
Timeout 600

Then stop and restart the OMS in order for the new settings to take effect. This will allow the system to take more time before dropping the agent connection (and causing it to retry again)


2.2.2 Installing Management Agents on Clusters

The agent installation does not discover cluster database targets out of the box. You should be able to perform the cluster database discovery after the agent installation using the Add Database Target wizard (Targets -> Databases -> Add).

2.2.3 Recovering the targets.xml file (Bug 3678671)

If you are configuring more than one Application Server (AS) home for central management by the same Oracle management agent at the same time, an internal file called targets.xml could be accidentally deleted.

In the event that this happens, you can recover all AS targets by following these steps:

Remove the centralagents.lst file if any, from the AS home (located under sysman/emd)
Create a new targets.xml (under sysman/emd of the agent's home directory) with the following entry:






where is the full hostname ex: my-sun.us.oracle.com

Reload the agent by running "emctl reload agent" from the agent's home directory
Reconfigure the AS home from Application Server Control's infrastructure tab
By doing so you will restore all AS homes to the internal file (targets.xml). However, any target that does not belong to the AS home will not be recovered. To recover those targets you will need to manually rediscover them.

2.2.4 Deleting a database target on the windows platform (Bug 3488516)

If you are deleting a database target from a centrally managed home using DBCA on a windows machine, your change may not be fully propagated to the Oracle Management Agent. To workaround this issue, the database target needs to be explicitly deleted from the management agent’s home using Grid Control

3. Monitoring System

3.1.0 User-Defined Metrics
If the name specified for the database User Defined Metric is more than 30 characters, the metric will not be evaluated properly. Please ensure that the metric name is at most 30 characters (Bug 3301697)

3.2.0 Metric Errors for User-Defined Metrics
If the evaluation of a user-defined metric results in an error, then there will be no return value associated with the metric in the User-Defined Metrics page. The user should go to the Errors subtab of the Alerts tab to get information about the metric error (Bug 3301699)

3.3.0 Alert remains for deleted User-Defined Metrics
If a user deletes a User-Defined Metric that is still in warning or critical alert, then the alert will remain. Make sure the User-Defined Metric is in 'clear' state before deleting it (Bug 3301704)

3.4.0 Agent Unreachable in the Availability page
If a target is currently in the 'Agent Unreachable' state, then in the Availability page for that target, this 'Agent Unreachable' state is shown in 'white' (i.e. no color) in the bar graph that represents the target’s availability over time. Typically, the Agent Unreachable state is reached if there are network problems between the Oracle Management Service and the Agent that is monitoring the target, making the true target status unavailable for display in the console. Once the target is out of Agent Unreachable state, then the 'Agent Unreachable' period will be replaced in the bar graph with the appropriate status of the target (up, down, etc) (Bug 3152217)

3.5.0 Response Actions
3.5.1 Response Actions will not work for server-generated alerts. (Bug 3129353)
3.5.2 Response Actions for metrics can only be specified in the Console by Enterprise Manager users with superuser privileges
3.5.3. When specifying a Response Action script for the Target Down condition, you will need to be careful when saving your entry. Specifically, after you enter the response action script, click on the 'OK' button. This will bring you back to the Edit Thresholds page. From this page, click on the Cancel button, not the OK button. (If you click on the OK button, your response action script entry may not have been saved (Bug 3465495)). However, clicking on the Cancel button may result in not saving any other changes to the Edit Thresholds page. Hence, if you are making changes other than specifying a response action script, make sure you save these other changes first (by clicking on the OK button in the Edit Thresholds page), then enter in the Response Action script afterwards as indicated.

3.6.0 Issues with host metrics on NT 4.0

If you are running the agent on a Windows NT 4.0 host with more than 2 GB of RAM, you may find that certain host metrics related to host memory are inaccurate. (Bug 3294952)

3.7.0 Missing Help for Metric Details

In some cases when you are displaying a metric on the Metric Detail Page and you click Help, the help system displays the Table of Contents for the help system, not the appropriate help topic for the metric. In this situation, the workaround is to choose one of the Real Time options in the View Data field on the Metric Detail page, then click the Help link again. Then the help system will display the appropriate metric help topic. (Bug 3341196)

3.8.0 Host Metrics on Solaris 9

On Solaris 9, the Host target has been enhanced by the addition of a new metric - Aggregate resource usage statistics, gathered by User and by Project. These metrics are collected by using the "CIM Object Manager" on the Solaris 9 host. In order to get this information, the CIM Object Manager needs to be supplied with Username and Password of a user on that host so that authentication can successfully happen. These values can be provided by following the Monitoring Configuration link from the Host homepage for this target. (Bug 3330619)

The following are examples of Errors/Messages seen when the credentials have not been provided:

2004-01-01 19:09:53 Thread-1690 ERROR fetchlets.wbem: Server reported error.


2004-01-01 19:09:53 Thread-1690 ERROR fetchlets.wbem: Server reported error.
2004-01-01 19:09:53 Thread-1690 ERROR engine: [host,dlsun454.us.oracle.com,ProjectResourceUsage] : meegd_GetMetricData failed : Server reported error.
2004-01-01 19:10:53 Thread-1714 ERROR fetchlets.wbem: No username specified for WBEM fetchlet.
2004-01-01 19:10:53 Thread-1714 ERROR engine: [host,dlsun454.us.oracle.com,UserResourceUsage] : nmeegd_GetMetricData failed : No username specified for WBEM fetchlet.

3.9.0 Failure to Render Pages that Contain Real-time Metric Value

You may see errors such as: "An internal error has occurred" or "Error getting data for target {target-name}. Exception: null" when EM is trying to display a page that is showing real-time metrics for a target that was deleted and recreated. This may occur if you discover a target such as an application server, then remove it and rediscover it. If you see these errors, stop and restart the agent that is monitoring the target:

{agent-home}/bin/emctl stop agent
{agent-home}/bin/emctl start agent

(Bug 3381701)

4. Blackouts

4.1.0 Users Must Have Operator Privileges To Blackout Group
If a user creates a group and would later like to blackout the group, he should ask an Enterprise Manager superuser to grant him 'Operator' privileges on the group. Operator privileges are required to blackout a group and by default, 'Operator' privileges for a group are not automatically granted to the creator of a group.

4.2.0 Blackouts With Immediate Start Times May Be Delayed
The start of an immediate, one-time blackout or the first occurrence of recurring blackout with an immediate start time may be postponed by up to 2 minutes. The goal is to use the target's time as the blackout start time, and the target's time, in turn, is determined based on its monitoring Agent's last heartbeat time to the Oracle Management Service (OMS). The additional 2 minutes is added by the OMS to the Agent's heartbeat time in order to accommodate any network delays in receiving such heartbeat information. This time is then used as the blackout start time. This calculation was specifically designed to ensure the proper handling of potential metric severities pending from the Agent. Since blackouts suspend monitoring, severities that have a timestamp before the official blackout start time will be kept; otherwise, they will be discarded (Bug 3299198)

4.3.0 Agent Unreachable or Status Pending State
After a blackout on a target ends, Enterprise Manager re-evaluates the target status. During this brief evaluation period, users may see the target in either 'Agent Unreachable' state (for host targets) or 'Status Pending' state (for non-host targets). These states are temporary and will be resolved once the true status of the target is determined (Bug 3317976)

4.4.0 Operator Privileges Required to Blackout a Target
An Enterprise Manager administrator needs at least Operator privileges on a target in order to blackout the target. The Create Blackout UI will only display targets for which the current Enterprise Manager administrator has at least Operator privileges. When creating a blackout, if no targets are available for selection, this means the administrator does not have Operator privileges on any target. To resolve this, he should ask an Enterprise Manager super administrator (i.e. administrator with superuser privileges) to grant him Operator privileges on the targets that he wants to blackout.

4.5.0 Reason Field Required When Creating Blackout
When creating a blackout, a 'reason' for a blackout is a required field. Only Enterprise Manager administrators with superuser privileges (aka super administrators) can create a 'reason' during blackout creation. When creating a blackout, if the appropriate value for 'reason' is not available for selection, ask your Enterprise Manager super administrator to create the reason by creating a 'dummy' blackout that has the appropriate reason in it. The dummy blackout can be, for example, a future blackout for any target. As soon as the dummy blackout is created, the reason has also been created. At that point, he should STOP and DELETE the dummy blackout. Make sure the dummy blackout DOES NOT ACTUALLY START and is DELETED because a blackout will stop monitoring the targets under blackout and will appear in that target's availability records. (ER 3145500)

4.6.0 Daylight Savings Time Not Supported
Daylight savings time is not supported for blackouts. Users need to manually adjust a blackout's schedule, as needed, to accommodate schedule changes due to daylight savings time.

4.7.0 Availability History
When a blackout is submitted and the Oracle Management Service is down for the entire duration of the blackout, and no severities come in for a target during the blackout window, the blackout window will be compressed to a one-second window in the availability history. Note that this does not cause any unnecessary notifications to be sent, since it only happens when no severities are generated for the target during the blackout. (Bug 3317965)

5. Jobs

5.1.0 Deleting job runs
If a user attempts to delete a job run wherein at least one execution in the job run is still active (i.e. Scheduled, Running, or in Suspended status), then he may see this error message: “ The specified job, job run or execution is still active. It must finish running, or be stopped before it can be deleted. Filter on status 'Active' to see active executions.”. 'Active' refers to a combination of the following status values: Scheduled, Running, or Suspended. However, ‘Active’ is not available as one of the search filter for the Status search field. Hence to see all active job executions, you can perform separate searches on jobs where the Status is Scheduled, Running, or Suspended (Bug 3256113)

5.2.0 Host Command Against Groups
If a user submits a Host Command job against a Group of different target types, then the Host Command job will automatically run against the Host targets in the group. In order to run the Host command against targets of a different type (e.g. databases), the user can choose those targets directly, or in the case of database targets, choose a Database Group containing those databases (Bug 3241432)

5.3.0 RefreshFromMetalink Jobs
When querying jobs, if a user chooses the job type 'RefreshFromMetalink', then, for convenience, Enterprise Manager automatically assumes the value of the Target Type filter is 'System' (regardless of what the user has selected for 'TargetType'). The RefreshFromMetalink jobs always have 'System' as their associated target type, and thus will only be found if the Target Type is set to 'System' (Bug 3230111)

5.4.0 CloneHome and Patch Jobs Are Not Supported In The Library
In the Job Library page, the 'CloneHome' and 'Patch' jobs are listed as job types for which library jobs can be created. However, these job types cannot be saved to the library, they can only be created and submitted as active jobs (Bug 3281143)

5.5.0 Credentials for Host Command Jobs Against Database Targets
A user's preferred credentials for a database target consist of normal database credentials, privileged database credentials, and host credentials. When creating a Host Command job against a database target, the Create Job UI allows the user to select 'Normal' or 'Privileged' preferred credentials. Either option will result in using the same host credentials -- these are the single host credentials that are part of the database target credentials (Bug3290724)

5.6.0 Single group target in Host Command/SQL Script jobs
When creating a Host Command or SQL Script job, if any type of group is selected as a target for the job, then no other targets can be added to the job.

5.7.0 Immediate, Repeating Jobs
When specifying the schedule for a job, if 'Immediately' is selected for the Start time, then any other repeat option specified will be ignored. Hence, the job will become an immediate, one-time job. If the user wants to create a job that runs immediately and also repeats at regular intervals, then he should choose the 'Later' Start option, and specify a time a few minutes after the current time. These additional minutes are required to accommodate the time needed to process and submit the job from the Oracle Management Service to the agents (Bug 3262332)

5.8.0 Editing Finished Jobs
When attempting to edit a job, a user might see this error message: "No scheduled executions - this job cannot be modified". This means that based on the job's schedule, Enterprise Manager has determined that there are no future job executions scheduled for this job. Hence editing the parameters of the job does not make sense and thus will be prevented. However, the access page of the job can always be modified by the job owner, even after the job finishes, in order to share job output with others.

5.9.0 Searching For Jobs Against Groups
If the user uses the 'Target Type' filter to search for jobs against groups, make sure the Target Type filter is set to one of the following: 'All Groups', 'Database Group', 'Group' or 'Host Group'. Setting the filter to 'All Targets' will not work in this case because 'All Targets' refers to individual non-group targets (Bug 3282349)

5.10.0 Skipped Executions
For recurring jobs, a future execution of a job is only scheduled when the current execution is completed. There may be cases wherein some job executions are skipped. The most common causes are:
a) the currently running execution is taking longer than expected, and continues to run during the time the next execution(s) is supposed to run

b) the current execution has been suspended by the user, and continues to be suspended during the time the next execution(s) is supposed to run

c) the current execution has been suspended by the system (due to the agent becoming unreachable or the target of the job being under blackout), and continues to be suspended during the time the next execution(s) is supposed to run

When the current execution finally completes, the next job execution is scheduled. The next scheduled execution is always a date in the future, hence any executions that were supposed to run in the past but were skipped will never be run. The Enterprise Manager console will NOT display these skipped executions. If some job executions seem to be missing from the jobs table in the console, these are most likely due to the cases enumerated above. (ER 3330400)

5.11.0 Oracle Scheduler Jobs vs Enterprise Manager Jobs
Users have access to 2 types of job systems: Oracle Scheduler jobs and Enterprise Manager jobs. The former are jobs that run inside the database; the latter are jobs that run outside the database using the Oracle Management Agent. You can access the Oracle Scheduler jobs using the 'Jobs' link under the Scheduler section in the database Administration page. You can access the Enterprise Manager jobs from the Jobs primary tab in the console, or by using the 'Jobs' link under the Related Links section of any database page: Home page, Performance, Administration, Maintenance. Also, the Job Activity section in any target home page refers to Enterprise Manager job activity (Bug 3312974)

5.12.0 Daylight Savings Time Not Supported
Daylight savings time is not supported for jobs. Users need to manually adjust a job's schedule, as needed, to accommodate schedule changes due to daylight savings time. (Bug 3328590)

5.13.0 Secure Agents
We strongly recommend that all Grid Control agents be secured on install. Leaving unsecured agents in a deployment can cause problems especially if you run clone jobs across agents, and only one of the two agents is secured. In such a scenario, the clone will always fail.

6. Database Management

6.1.0 Data Guard

6.1.1 Metric Collection Errors in Standby Databases
After switching a primary database to a standby database you may notice a few metric collection errors for the new standby database. These errors can be ignored. The workaround to clear the metric collection errors is to briefly open the database in read-only mode. (Bug 3312518)

6.1.2 Data Guard: Upgrading Protection Mode Causes Exception With a RAC Database
When upgrading the protection mode in a Data Guard configuration that has a RAC database, connecting to the RAC database at the end of a successful protection mode upgrade may throw an exception. The workaround is to log out of the database and log back in again from Enterprise Manager. (Bug 3345796)

6.1.3 Using Primary Database as Recovery Area for Locally-Archived Logs
If your primary database is using a recovery area for locally-archived logs, ensure that the log_archive_dest_10 parameter is set to the following before creating a physical or logical standby database:

log_archive_dest_10='location=use_db_recovery_file_dest','valid_for=(ONLINE_LOGFILE, ALL_ROLES)'

If this parameter is not set on a primary database that uses a recovery area before switching over to a logical standby database, online logfile archival will fail with an ORA-16104 after the database is activated as a logical standby. If you neglect to set this parameter prior to switchover and see this problem, it can be fixed by setting the above parameter on the logical standby database. (Bug 3349731)

6.1.4 Creating A Standby Database For A Cluster Database (RAC) Primary Database
When creating standby databases for cluster database (RAC) primary databases in Enterprise Manager you must follow all the configuration steps described in Section 2.2 ("Setting Up the Broker Configuration Files") of the Data Guard Broker book in the Oracle Database 10g documentation.

6.1.5 Standby Database Creation Error on Windows
On the Windows platform only, if the name (including the path) of a standby datafile or logfile contains the characters "p" an error will result.
If this occurs in the directory component of a name, either rename the directory and change the standby file names accordingly, or choose a completely different directory to place the files into, and rename the files accordingly using the Customize button on the File Locations page of the Add Standby wizard.
If the "p" occurs in the file name component of the name, rather than the directory component, change the filename using the Customize button on the File Locations page of the Add Standby wizard. (Bug 3655231)

6.2 Reorganize Objects

6.2.1 The database can be adversely affected if metadata changes occur between the time the reorganization script is generated and the time it is executed.

6.2.2 The database can be adversely affected if an offline reorganization is attempted while DDL is in progress against the tables being reorganized.

6.2.3 Before using the Reorganize Objects wizard or the Make Tablespace Locally Managed wizard Oracle recommends that you ensure you have a very recent backup of your database.

6.2.4 For 10g database targets only, attempting to do an online reorganization of the outer join table of a bitmap join index will fail. Use the offline method for reorganizing such a table or reorganizing a tablespace containing such a table. (Bug 3330361)

6.2.5 For 10g database targets only, attempting to do an online reorganization of a tablespace that contains an index-organized table overflow segment partition will fail when the "tablespace rename" option is used for the Scratch Tablespace. Use the offline method for reorganizing such a tablespace, or specify a scratch tablespace instead of using the tablespace rename option (Bug 3334289)

6.2.6 Error in Impact Report
When reorganizing a tablespace in a 10g database, you may encounter the following error message in the Impact Report:

This tablespace contains recycle-bin segments from tables that have been dropped. It cannot be reorganized. Use the PURGE TABLESPACE command to remove the recycle-bin segments from the tablespace and re-try the reorganization.
The PURGE TABLESPACE command only removes recycle-bin segments belonging to the currently connected user. Therefore, it may not remove all the recycle-bin segments in the tablespace. You can determine which users have recycle-bin segments in a target tablespace using the following query:

SELECT DISTINCT owner FROM dba_recyclebin WHERE ts_name = 'TARGET_TS';

You can then use the following command to purge the segments for each of the users:

PURGE TABLESPACE target_ts USER scott;


An alternative is to use the following command:

PURGE DBA_RECYCLEBIN;

This can be used only if you have SYSDBA system privilege. It removes all objects from the recycle bin, regardless of user.

6.3.0 Net Services Administration
For the 'Directory Naming' feature to work, a valid ldap.ora file must exist in the 'Configuration File Location' selected by the user in the 'Net Services Administration' page. The ldap.ora file should be created using NetCA or OidCA. If the ldap.ora file location is not OH etworkadmin or OHldapadmin, then you must set the TNS_ADMIN / LDAP_ADMIN environment variable to that location before starting Database Control.

6.4.0 Shutting Down the Database
Shutting down a database in normal mode does not work in Enterprise Manager. However, a database can be shut down in transactional, immediate and abort modes with Enterprise Manager. Use SQL*Plus to shut down a database in normal mode. (Bug 3239095)

6.5.0 Running SQL Access Advisor With Materialized View Recommendation May Cause Library Cache Deadlock
A deadlock on the library cache may happen when a select statement is evaluated for rewrite at the same time as a create materialized view statement requests exclusive locks on its base tables. This could happen when a SQL Access Advisor is running to generate materialized view recommendations while another user issues a select statement at the same time that references the same base tables. The workaround is to stop the SQL Access Advisor job and the deadlock will be released. (RDBMS bug 3230878, 2018081, 2124800)

6.6.0 10g RAC Database Recovery in Archivelog Mode
When a 10g cluster database is in archivelog mode, in order to perform a restore and recovery of the entire database, the user should only leave one RAC instance up and shut down all other RAC instances. Affected areas include whole database complete/point-in-time/flashback recovery. (Bug 3334212)

6.7.0 Default Database Credentials May Not Be Used for Cluster Databases
Default database credentials will not be picked up for cluster databases unless at least one cluster database was already discovered at the time when the default credentials were set. The workaround for this problem is to delete and reset the default database credentials. (Bug 3312344)

6.8.0 Launch iSQL*Plus Requires Configuration
For Grid Control, the iSQL*Plus server is not shipped with the OMS (Oracle Management Services). Therefore iSQL*Plus needs to be manually configured to point to any running Oracle Database 10g iSQL *Plus server. All the target databases discovered in this Grid Control application will then use this iSQL*Plus server for their queries.

To configure iSQL*Plus, two entries in the $ORACLE_HOME/sysman/config/emoms.properties file on the OMS (Oracle Management Server) machine needs to be populated by replacing the following placeholders: replace %ISQLPLUS_PROTOCOL%, %ISQLPLUS_HOST% and %ISQLPLUS_PORT% with the protocol, the host and the port of the iSQL*Plus server you want to use.

For example:
oracle.sysman.db.isqlplusUrl=http://rooster.au.oracle.com:8888/isqlplus/dynamic
oracle.sysman.db.isqlplusWebDBAUrl=http://rooster.au.oracle.com:8888/isqlplus/dba/dynamic

For RAC applications, the emoms.properties file for all nodes of the cluster must also have these two entries configured. After configuring the entries, the OMS needs to be restarted to pick up the change.

6.9.0 Editing Objects
Enterprise Manager only supports having one browser window open for editing an object. For example, Enterprise Manager only supports editing one tablespace at a time. Note, if a window is read only, you may have multiple browser windows open.

6.10.0 Target Names For Oracle Suggested Backup Strategy
When Oracle's Suggested Backup Strategy is to be used for offline backups, the target name should be less than 52 characters in length. (Bug 3347338)

6.11.0 Viewing of SQL Execution Plan in Korean
There is a known bug where viewing an execution plan of a Period SQL causes an internal server error. This is a bug that is unique to Korean and does not reproduce in Japanese or Chinese. The only workaround currently available is to run the product in a non-Korean language when you need to view this page. (Bug 3380531)

6.12.0 Garbled Characters Shown in "PGA Memory Usage Details" Page
The "PGA Memory Usage Details" page under Administration -> Instance:Memory Parameters -> PGA will show garbled characters when the product is run in a multi-byte (Asian) language environment. The only workaround currently available is to run the product in a single-byte Western European language when you need to view this page. (Bug 3272773)

6.13.0 Database Performance Page Shows Grey Area Instead of Graphs in Internet Explorer
Sometimes in Internet Explorer the SVG graphs on the Database Performance page don't render. Instead the page displays a large grey area with no graphs at all.
To fix this problem change a setting in your Internet Explorer browser as follows:
Tools -> Internet Options -> Advanced
Scroll down to Security and uncheck "Do Not Save Encrypted Pages to Disk".
This problem does not apply when SVG has not been enabled, i.e. you are viewing the graphs in GIF mode.

7.1.0 Oracle Targets Missing From Deployments Summary Rollup Tables
If Oracle software was not installed using the Oracle Universal Installer or was not cloned using Enterprise Manager, the software will not show up in Deployments Summary rollup tables. This is due to the fact that data is collected from the Oracle Universal Installer inventory. Software not installed via Oracle Universal Installer will not have an entry in the inventory, and as a result cannot be displayed in rollup tables.

7.2.0 Cloning Support For Legacy Deployments Not Available At Release Time For All Platforms
Cloning for legacy software deployments on certain platforms may not be immediately available at the time of
release. The Cloning wizard will detect this and point you to the appropriate web site to go check for the availability of an update.

7.3.0 Agent Patching On RAC Is Not Supported
Patching of the Enterprise Manager Agent in a RAC setup is not currently supported, use the command line utilities instead.

7.4.0 Patch Number Inconsistencies
On the Patch: Select Patch page, you can specify a MetaLink patch number to locate a patch published at the
OracleMetaLink web site. For some interim patches, a second patch number is used that is unique to the platform. In such cases, this number will appear in the readme as the interim patch number, e.g.: "Interim Patch:3243133 Base Bug:3239222". This is the number that is returned as part of enterprise configuration searches and is the number that will appear on the Oracle Home page. (Bug 3320868)

7.5.0 Cloning Of Oracle Home Fails If A 9.2 Agent Is Running
If a 9.2 Enterprise Manager Agent is running on a host, cloning operations will fail. You must stop the Enterprise
Manager 9.2 Agent before performing any clone operation for it to succeed. (Bug 3164441)

7.6.0 Oracle Home Link In Critical Patch Advisory Page Is Broken If The Name Contains An Underscore
The Critical Patch Advisory page entitled: "Policy Rule: Critical Patch Advisories for Oracle Homes" will have
broken Oracle Home links in it's table, for any Oracle Homes that contain underscores ("_") in their name (Bug
3324506)

7.7.0 MetaLink Password Change May Have Delayed Propagation
Password changes performed on the MetaLink site may have a slight propagation delay that affects Enterprise
Manager. If you get MetaLink credential errors in Enterprise Manager shortly after having changed your password on MetaLink, simply wait a while and retry later (Bugs 3319786, 3265575)

7.8.0 Deployments Summary Table Shows Database Targets That Aren't Directly Visible In The "Oracle Home" Page
The Deployments Summary table shows all database targets, whether they are standalone or part of another Oracle product, but if you drill down on a database target that is part of another Oracle product, and from there drill down into the "Oracle Home" page, you won't see the database component as part of that home, as it's in fact a subcomponent of another Oracle product. (Bug 3336562)

7.9.0 Database Patches Prior To 9.2.0.5 And 10g May Have Bugs Fixed Inconsistencies
Database patches prior to versions 9.2.0.5 and 10g may have inconsistencies with regards to the bugs fixed list, both within the readme and the metadata written by the patch itself.. Enterprise Manager uses the information written by the patch.

7.10.0 Patch Version Information Not Available In Enterprise Manager
Enterprise Manager does not currently have patch version information in it's repository. Oracle may re-release a
patch without updating the identification number used by Enterprise Manager, and thus it will not be possible to
discern the old from the new.

7.11.0 Critical Patch Advisory Message Incorrectly Uses Term Patchset Instead of Version
When listing all possible patches that can be used to address a Critical Patch Advisory, Enterprise Manager sometimes lists a patch and the version number to which the patch applies, for instance Alert 46 can be fixed with an interim patch for 9201 or an interim patch for 9202. In some cases, the phrase “(Requires patchset x.y.z.a)” is included to make clear which version of the patch is being referred to. This may lead to confusion when the patch is for the initial release of the product, e.g. you may see “Requires patchset 9.2.0.1”. In such cases, there is no patchset – it simply refers to the patch for version 9.2.0.1. (Bug 3347848)

7.12.0 Availability of Critical Patch Advisories
At the time of this release, Critical Patch Advisories are being provided for the Oracle database server (versions 9.2 and higher) and for iAS (versions 9.0.4 and higher). Advisories for other products may be added at any time. The following Operating System platforms are currently supported:

AIX5L Based Systems (64-bit)
HP Tru64 UNIX
HP-UX Itaninum -- HP-UX PA-RISC (64-bit)
IBM zSeries Based Linux
Linux Itaninum
Linux x86
Microsoft Windows (32-bit)
Microsoft Windows (64-bit)
Solaris Operating System (SPARC 32-bit)
Solaris Operating System (SPARC 64-bit) .
Solaris Operating System (x86)

7.13.0 Non-Oracle Database 10g Home Clonerstage Availability
For Oracle Database homes prior to 10g, as well as Oracle Application Server homes 9.0.4, Enterprise Manager requires a clonerstage to perform home cloning operations, as this functionality wasn't present in the Oracle Universal Installer for products shipped prior to Oracle Database 10g. Only a limited amount of clonerstages are shipped with the product, and thus when you attempt to clone a home for which a clonerstage isn't currently available within your Oracle Management Service, you'll be presented with a page telling you so and pointing you to the appropriate location to download that clonerstage.

7.14.0 Home Cloning for Homes Requiring Clonerstages Absolute Path Issues When using Multiple Oracle Management Services
If you have multiple Oracle Management Services within your Grid Control setup and you attempt to perform a home cloning operation on a legacy home that requires a clonerstage, the path to that clonerstage is an absolute one, and thus if the job runs on a different Oracle Management Service than the one that submitted the job and if the path on both Oracle Management Services is different, the job will fail. To avoid this problem, either use consistent paths on your multiple Oracle Management Services or as an alternative for non-Windows platforms, simply create appropriate symbolic links so that all path combinations are valid on all Oracle Management Services. (Bug 3349916)

7.15.0 Installing Home In Pre-Existing Path Prevents Later Home Cloning
If you install Oracle Database 10g in a path that already exists, a clone home operation on that home at a later time will fail. To avoid this, simply ensure that the path you're installing in (ORACLE_HOME) doesn't already exist, letting the installer take care of creating it for you. If you do install into a pre-existing directory, you may see a later error from the home cloning wizard, that mentions the install.platform file could not be read. To work around this problem, copy the install.platform file from the central OUI inventory directory into the home you are attempting to clone.(Bug 3235500)

7.16.0 Manually Uploading Large Patches to the Patch Cache Fails
Manually uploading a patch larger than 128 MB to the patch cache will fail. However the normal way of deploying patches from Grid Control, which involves an automatic download from Metalink of the patch is not affected by this problem. (Bug 3719280)

7.17.0 Job Failure when Cloning Homes

The CloneHome job may fail during the updateBuiltInClonable step. If this happens, stop and restart the agent on the node you are cloning. (Bug 3400612)


7.18.0 Content of Search Results Saved To Local .csv File Are Unreadable

The list of Oracle Products Installed in Oracle Homes can be stored to a local .csv file. This feature can be found under Deployment-> Search -> "Oracle Products Installed in Oracle Homes"-> "Save to File". The content of this file is stored using the UTF-8 character set, which may not be recognized by some desktop applications. If you find garbled characters in this file, try opening it with an application that can handle the UTF-8 character set, such as most web browsers. (Bug 3184524)


7.19.0 Files to Exclude When Cloning an Oracle Home

If you clone an Oracle Database 10g home, you should exclude some additional files from being copied to the new home. To do this, add the following to the home cloning wizard's "Files to Exclude" list: listener.ora, tnsnames.ora, sqlnet.ora. If these files have already been copied during home cloning, delete them from the new home before running the DB Clone utility. (Bug 3394291)


7.20.0 Database Cloning on Windows Requires A User Of The DBA Group If Not Logged in As SysDBA
In order to successfully perform a database cloning operation on a Windows target platform, you either have to be logged into the database as SYSDBA, or performing the operation on the source target as an OS user who's a member of the DBA group. If you don't meet either one of these two requirements, the clone operation will fail. (Bug 3684746)


7.21.0 Database Cloning Using Shared Directory On Windows Target Fails
If you attempt to use the "Use the files directly from a shared directory" option whe

来自 “ ITPUB博客 ” ,链接:http://blog.itpub.net/7318139/viewspace-969837/,如需转载,请注明出处,否则将追究法律责任。

请登录后发表评论 登录
全部评论