Table Of Contents
Release Notes for
Cisco Element Management Framework v3.1Introduction of Versioned Shared Libraries
Known Limitations in the Release
Cisco EMF CORBA Gateway Runtime component
How to remove the Orbix demon from Cisco EMF process control
Resolved Between Cisco EMF v3.1 EFT and This Release
Caveats Resolved By Cisco EMF v3.0 Patches Applied Between Cisco EMF v3.1 EFT and This Release
Bugs Fixed in CEMF 3.0 Patch 7
Bugs Fixed in CEMF 3.0 Patch 6
Bugs Fixed in CEMF 3.0 Patch 5
Bugs Fixed in CEMF 3.0 Patch 4
Release Notes for
Cisco Element Management Framework v3.1
Contents
This document is divided into the following sections:
•
"System Requirements" section
•
•
•
•
•
•
Introduction
Cisco Element Management Framework (Cisco EMF) is the homogeneous element management layer of the Cisco Internet Operation Support System (OSS) building on and adding to the Cisco EMF v3.0 product functionality. It comprises a highly scalable framework designed to support carrier-class Element Managers (EMs) across Cisco's service-provider product lines with network event management. This flexible, modular model allows Service Providers to install the mix of Element Managers they need to support their dynamic businesses. It also enables rapid development and deployment of new Element Managers, permitting Service Providers to more rapidly introduce and manage new services.
Cisco EMF provides common interfaces and element management services to applications in the network and service management levels of Cisco Internet OSS. Cisco's EcoSystem partners can also build third-party element management systems to provide element management support in mixed network environments.
Cisco EMF v3.1 end customer functionality focuses on integrating and bundling the Event Manager application and improving the northbound event interfaces supported by the framework to further enhance integration of Cisco EMF into customers' OSS environments.
Northbound Event Interfaces
Forwarding of Cisco EMF Events encapsulated in SNMP Traps to any SNMP Network Management System.
For example, with HP-OEMF Version 4.0, the information on both the managed network and the events occurring within the managed network is passed northbound to the HP-OEMF umbrella fault management system, and the managed network in the form of Cisco EMF Object Model is exported via a flat file.
CORBA Gateway
CORBA Gateway v2.0 introduces support for a CORBA event feed to enable asynchronous event notification from Element Managers to external CORBA clients. The event feed supports the ability to specify interest registration, via an appropriate filtering mechanism.
Event Manager
The Event Manager is integrated as a constituent part of Cisco EMF v3.1 functionality. The main benefits of the Event Manager functions are:
•
•
•
Introduction of Versioned Shared Libraries
In addition Cisco EMF v3.1 introduces versioned libraries. This is a key technology enhancement which significantly increases the ability for Cisco to make library patches and maintenance releases to the framework where previously a full, minor release would have been required. This directly improves the longevity and stability of deployed software, for example allowing framework bugs that are affecting an Element Manager to be patched live without impacting other Element Managers deployed on the same platform.
System Requirements
Cisco EMF can be deployed in a number of different configurations, from a single hardware system to a multi-site management system, using Sun Solaris/SPARC hardware with Solaris, Release 2.6.
Software Compatibility
Cisco Element Management Framework v3.1 is compatible with the following:
Cisco EMF-Software Development Toolkit v1.3
Cisco EMF CORBA Gateway Developer Toolkit v2.0
Enhancements in this Release
1.
2.
–
–
–
–
–
Note
3.
4.
5.
6.
Backup Mechanism
The backup mechanism has been modified in Cisco EMF 3.1 to ensure that database integrity is maintained for backup and restore. This is not the case in releases prior to Cisco EMF 3.0 Patch 9.
Before a backup commences, Cisco EMF locks deployment, alarm processing, installation and changes to the object model. They are unlocked when backup ends.
Messages relating to this are displayed while a backup is in progress.
During a backup, the user will not be able:
–
–
–
–
Known issues
1.
Patching and backing up must be mutually exclusive to avoid configuration file inconsistencies in the restored system. Therefore, we advise to avoid patching Cisco EMF around times of backup.
2.
Object groups can be modified by the GUI while backup is in progress so the user may need to repopulate the Object Group when databases are restored.
Future enhancements
In this release, Cisco EMF fails attempts to provision when backup is in progress. In Cisco EMF 3.1 Patch 1, Cisco EMF will buffer new requests until it is safe to resume.
Known Limitations in the Release
Cisco EMF CORBA Gateway Runtime component
Iona Orbix issue : If Cisco EMF is installed on a Sun workstation which is not running Domain Naming Service (DNS), the Cisco EMF-CORBA Gateway Runtime Server fails to start.
Workaround: either
re-install Cisco EMF, and when asked for a domain press Return, this enters an empty string. Do not re-install Orbix
or
install Cisco EMF only on workstations which run DNS.
How to remove the Orbix demon from Cisco EMF process control
Some clients may wish the CORBA Orbix demon to run constantly even after Cisco EMF has shutdown. To do this the client must execute the following commands:
> cd <CEMF_ROOT>
> mkdir backupProcessFile
> cd config/processes/
> mv S08orbixd_3.0.x /<CEMF_ROOT>/backupProcessFile/.
> cd <CEMF_ROOT>/bin
> sysmgrClient -k orbixd
> <CEMF_ROOT>/Orbix/bin/orbixd -s &
where <CEMF_ROOT> is substituted for the current installation directory.
Installation
The Cisco Element Management Framework Installation and Administration Guide Version v3.1, Part Number 78-10792-01, provides information to install, license, and configure the product and includes detailed administration tasks.
Important Notes
Default Backup Directory
If the default installation settings were used to install v3.0, the default backup directory has changed between versions as follows:
/opt/AVBackup
/opt/Backup
Desktop Color Setup
Cisco EMF uses the color setup of your display. If your system is set up to display text in white, this may give problems in some Cisco EMF applications, for example on an event group icon in the Event Groups application. To avoid this, select the default color configuration.
Caveats
Note
Note
Open Caveats
•
Problem: The path is not getting updated while adding objects.
Cisco EMF provides the feature to allow more than one reference to an object to exist in the same view. For example:
Physical:/Site-1/object1
Physical:/Region-2/object1
Some applications, for example the Map Viewer, will display the path of the current object being displayed in the title bar. This is often the first containment path which can be found in the current view.
The fact that a new path which is given to an object, for example by adding it beneath another object, is not displayed in the Map Viewers title does not mean that the path does not exist.
Workaround: This can be checked by expanding the view on the left hand side of the Map Viewer to find the references to an object.
•
Problem: ./avload -remove does not clear profileContainment for Element Managers.
It is possible to design Element Manager dialogs which allow profiles to be created and used. Profiles in this context are pre-defined values for a number of attributes, which can be applied to one or more objects.
If such profiles are created using an Element Manager dialog, and the Element Manager is removed from Cisco EMF, the profiles will not be removed from the system.
When the same Element Manager is subsequently reinstalled, the profiles from before may be visible again, though they may not operate correctly.
Workaround: The workaround is to remove all profiles manually, by using the Delete Profile option on the Edit menu in the relevant dialogs, prior to normal removal of the Element Manager.
•
Problem: Selection of Objects when Invoking a Controller Action
Element Manager dialogs can be designed using Cisco EMF-SDK to allow actions to be launched on particular objects, by presenting a button for the user to press. In some cases, the action may be a scheduled one, in which case the user is prompted for the time at which the action should be carried out.
Once the correct date and time have been entered, the objects upon which the action will be invoked will be the objects selected in the Element Manager dialog at that time. If the user changes the object selection after the schedule dialog appears, then the new selection of objects will be used.
Workaround: None
•
Problem: Auto Discovery and Changing IP Addresses
If you need to change the IP Address of a managed object in Cisco EMF, special care is required to ensure the correct operation of Auto Discovery. The reasons for this are described briefly below, along with simple steps which can be taken to change addresses consistently.
When a managed object in Cisco EMF has an IP Address, this information is normally stored as an attribute of the object. However, Auto Discovery also records this IP Address by using it for the object name inside the Network view.
If you need to change the IP Address of an object which Auto Discovery is aware of, then it is also important to rename the object in the Network view. When you attempt to change the IP Address using Object Configuration, a warning to this effect is given.
Similarly, if a user renames an object in the Network view, this should only be done because the IP Address has changed. The new name should be the new IP Address of the object, and this address should also be set using Object Configuration. When a user attempts to rename an object in the Network view, a warning to this effect is given.
If the IP Address of an object is only changed by setting the IP Address attribute using Object Configuration, then the following will occur:
a.
b.
If an object is renamed to a new address within the Network view, without changing the IP Address attribute, then any attributes which are retrieved in respect of the object which rely on the IP Address will continue to be retrieved using the old address, contrary to what a user might expect by looking in the Network view.
Workaround: To change the IP Address of a managed object while avoiding the above is simple:
Step 1
Step 2
The following documentation may be useful, in the section Managing Objects in Cisco EMF in the Cisco EMF v3.1 User Guide and Help system:
–
–
Note
Note
•
Problem: User can not move up the container hierarchy while focus is on an object itself
In the Map Viewer application, the up arrow key can be used to move the cursor up the View representation on the left hand side.
However, if the keyboard focus is on the Map on the right hand side, the up arrow key will do nothing.
Workaround: To show a map for a higher level within the view, the appropriate object should be selected on the left hand side, or the up arrow on the toolbar should be clicked using the mouse.
•
Problem: Upgrade Unix EM produces two options for av install -r - DEP. This occurs in the case when the Unix Manager Element Manager is installed on Cisco EMF and an upgrade for this Element Manager is installed.
When trying to remove the Element Manager from the system two options will be selectable, the original version and the upgraded version, even though only the upgraded version is installed.
Workaround: Selecting either option for removal will successfully remove the Unix Manager Element Manager from the system.
•
Problem: Availability of Popup Menu Options in Applications
Some applications allow a popup menu to be shown for objects in the system, so that further applications can be launched.
In some cases, features will appear to be available from these menus, when they are not applicable to the user. However, security is maintained when the user actually selects an option, because at this point a check is made to ensure that the feature is genuinely available.
Workaround: None
•
Problem: discoverySchedule interval ignored if Cisco EMF is restarted. This means that Cisco EMF does not remember when the last scheduled discovery took place if Cisco EMF is stopped and restarted.
For example, this is observed when a discoverySchedule file is added with a StartTime of say 11:00 am; and a Period of 100 days. If this schedule has already been run in the previous day, and Cisco EMF is stopped (./cemf stop) and restarted, it will be rerun on the day Cisco EMF is restarted when 11:00 am rolls around.
Two significant consequences of this are:
a.
b.
Explanation: This is expected system behavior. The Auto Discovery application does not record when scheduled discoveries are run. All discovery scheduler timers are restarted following restarts and so discoveries will next to be run when their start time is reached.
If Cisco EMF is restarted, all scheduled discoveries will be run when their start time is next reached. This means if we have a discovery scheduled for 11am every seven days, and two days after this scheduled discovery has been run, Cisco EMF needs to be restarted, the next time this scheduled discovery will be run, will be at the next 11am reached and not in five days time.
Workaround: None
•
Problem: ./cemf session succeeds without any Cisco EMF processes running.
If the user attempts to start a session on a Cisco EMF Client installation, when the corresponding Server has not been started, the session will fail to open and no error will be displayed. It may also be necessary to cancel the operation using the CTRL-C interrupt sequence.
Workaround: The workaround is to ensure that the Server installation is running properly before attempting to start the session on the Client.
•
Problem: [CSCdp31915] accelerator keys in most of the places are not working
Use of Accelerator Keys in Applications
It is common for applications to allow some menu options to be selected by use of accelerator keys, for example, Ctrl+Q. In some applications within Cisco EMF, these short keys will fail to operate.
Workaround: In such cases, this is not a failure of the operation: simply choose the option from the correct menu using the mouse, and it will be available.
•
Problem: Parsers should check type of file before parsing
Symptom: Cisco EMF can fail to start properly or fail to de-install.
Condition: 'Rogue' files are put into directories that are parsed by Cisco EMF. Examples are core files, corrupt files and duplicate files. This condition is most likely to happen during development.
Workaround: Ensure that only the correct file types are in the assigned directories.
•
Problem: Object Group Manager hogs the processor when a query on all objects-DEP
Symptom: The system can become unresponsive after running a query.
Condition: The query specifies all objects in a view and all object types.
Workaround: Queries based on many types should not be performed on large scopes.
•
Problem: Alarms loading in acked state when they haven't been acked
The Event Browser application can be used to acknowledge events. Problems can occur if the same Cisco EMF user runs several sessions, each with an Event Browser opened on a different object. This can happen, for example, when several people log in as the admin user, Event Browsers in one session may appear to show events as acknowledged, as a result of events being acknowledge in a browser in another session, even though it may be on a different object. This does not occur when a different Cisco EMF user is logged in to each session.
Workaround: If the Event Browser is believed to show incorrect information, the refresh option can be selected to remedy this.
•
Problem: [CSCdr88128] site objects are allowed to be deployed on network tree
Symptom: Class-site objects can be incorrectly deployed via the Corba Gateway.
Condition: It is possible to incorrectly deploy an object of class site in the Network view through the Corba Gateway. It should not be possible to place objects of that class in the Network view.
However, at the moment it is the responsibility of the CORBA client to follow containment rules correctly in the absence of platform validation.
Workaround: There is currently no workaround for this.
•
Problem: Access Manager allows the user to delete the other users.
When a user has been given sufficient privileges to change other users' passwords in the Access Manager, then such a user will also be allowed to delete other users.
Workaround: Care must be taken when allowing this level of access, to ensure that users will only be deleted when this is really desired.
•
Problem: Possible core dump if you delete an alarm group while you edit it
This issue relates to editing an event group in the Event Groups application (available by selecting 'Event Groups' from the Cisco EMF Launch Pad).
While a user is editing the notification profiles assigned to an event group:
- if the user closes the event group editor
- or if the same user or another user deletes the same event group
Then the Event Groups application can exit if the Finish button is clicked on the Notification Profile Dialog that will still be open.
Workaround: The workaround is to click the Cancel button instead of the Finish button.
•
Problem: Get told the trigger condition has changed when it hasn't
This issue relates to editing an event group in the Event Groups application (available by selecting 'Event Groups' from the Cisco EMF Launch Pad).
When one user edits an event group, any other user who is editing the same event group (or the same user editing the event group with a second editor) will be informed of the changes that have been made to the event group.
No matter what change has been made, the second user is always informed that the trigger conditions have changed. For example, if user1 only edits the name, user2 is informed that the name and the trigger conditions have been changed.
The user should be aware that, in these circumstances, if they are informed the trigger conditions have changed, they may not have.
Workaround: None
•
Problem: [CSCdr57032] Auto discover SNMP community strings wrong
Symptom: The user is unable to save configuration changes to auto discovered devices.
Condition: During auto discovery, the system sets the read/write community strings to the first community string to which a device responds. By default, this will be public. The device will fail to save any configuration changes while the write community is set to this value.
Workaround: To avoid the condition, in the Auto Discovery dialog, remove the public community string and then add the private community string to the community list. Then, auto discovering as normal will set both community strings to private which should allow configuration changes to be made.
To resolve the condition, the write community can be set to private by using the relevant Object Configuration dialog for the device.
•
Problem: The command "./cemf install -o -fullhelp" does not work correctly.
Symptom: The commands "cemf help full" and "cemf install -o fullhelp" do not appear to work.
Condition: When running "cemf help full" or "cemf install -o fullhelp" an error is returned.
Workaround: The command "cemf install -o fullhelp" is erroneous and should not be used.
There is no workaround for command "cemf help full". However, standard help can be requested by running "cemf help".
•
Problem: 2 scheduled discoveries within same address range - neither takes place
Auto Discovery can read schedules from an initialisation file and carry out operations as a result.
However, when one of these scheduled operations fails, the resulting behavior is not defined.
Workaround: The workaround is to arrange that when multiple operations are scheduled, they are not dependent on any other operation having completed successfully, reducing the likelihood of failure.
•
Problem: deploymentWizard:- Problem error reporting when attrs fail
Symptom: If an object is deployed, and the deployment fails to set some attribute values, Cisco EMF will report this. However it may also report that the object could not be added to one or more views.
Conditions: If objects are deployed and one or more of the attributes on those objects fail to have their values set as part of this deployment, then the status message may also indicate that the object could not be added to one or more views.
Workaround: The misleading error message regarding view failure can be safely ignored.
•
Problem: Batching problems with ObjectFileParser - DEV
Symptom: Objects deployed via file can fail to be unloaded if deployment fails.
Conditions: /config/init/objectFileParser.ini has a "MaxPerContext" setting that defaults to 100. If more than this number of objects are loaded into Cisco EMF using the objectFileParser, and one or more of the specifications contains an error, then some of the previously loaded objects will not be unloaded.
Workaround: Ensure that object specification files loaded using objectFileParser contain less than 100 object specifications.
•
Problem: if corba client dies while processing context, part cycle will hang
Symptom: Participation does not proceed when a participant does not receive the participation context.
Conditions: If a controller (which is a participant) crashes while it has the participation context, participation will not proceed as the next participant will not receive the context.
Workaround: Cisco EMF clients (e.g. EMs) written to be participants in the Cisco EMF deployment process must return the deployment context to the participation cycle. This allows deployment to proceed to the next participant in the cycle. Cisco EMF will be enhanced in the future to take account of the case where a controller may crash during deployment, and to react accordingly.
•
Problem: Deployment wizard hangs when deploying 8000+ objects.
Symptom: Deploying or deleting more than 100 objects in the same deployment context may sometimes cause the operation to fail.
Conditions: Currently within Cisco EMF there is a limit of 100 objects that can be contained within a participation context. Exceeding this limit may result in a participant failing to de-serialize the context, and the deployment or deletion operation may fail as a result of this.
Workaround: Ensure that a participation context contains no more than 100 objects.
•
Problem: Installation appears to attempt license key copy even if answered "n".
Symptom: During Cisco EMF installation, if the response to the question "Do you have a valid license key?" is "n", a message similar to the following will be seen:
AccessVision License Installer... Installing /opt/cemf/config/licenses/* Cannot find license file /opt/cemf/config/licenses/*
Conditions: Cisco EMF installation is being performed and no valid license key is provided by the person carrying out the installation.
Workaround: The message simply confirms that no license file was found, and that no license key has been copied. It can therefore be ignored.
•
Problem: class action checking in avload needs to be switched on
Symptom: If an error occurs during installation, the errors are correctly reported, but are followed by a misleading message saying that installation was successful.
Conditions: Cisco EMF installation is being performed and an error occurs.
Workaround: This message should be ignored if previous errors have been reported.
•
Problem: EM installation overwrites files, and could cause problems.-DEP
Symptom: Installing an EM may result in all of its configuration files not being parsed. Deinstalling the EM may then leave Cisco EMF in an inconsistent state.
Conditions: If an EM includes files of the same name as those present in the default Cisco EMF installation, they will not be re-parsed when the EM is installed. De-installing the the EM will remove these files, which could leave Cisco EMF in an inconsistent state.
Workaround: EMs should add new files where possible, rather than overwriting standard Cisco EMF files.
•
Problem: EM deinstallation unable to remove the Object Group
Symptom: When an Element Manager is de-installed from Cisco EMF, an error may be reported that an object group failed to be removed.
Conditions: If object groups are created as a result of installing an Element Manager, then de-installing the Element Manager will not remove those object groups. Cisco EMF will indicate that they could not be removed.
Workaround: This error can safely be ignored as it has no detrimental effect on the system. Object groups are not removed as part of an Element Manager de-installation, so that when the same Element Manager is installed again, then the object group will not need to be created as it already exists.
•
Problem: PAS could cause problems for OSTransactionControl users
Symptom: Client code can cause ObjectStore exceptions to be thrown.
Conditions: After retrieving a persistent pointer, client code may call a function that updates the PersistentAttributeStore instance. If afterwards, the client code does not request the persistent pointer again, but instead reuses the original address, the original address may now be invalid and so could cause the client code to throw an objectStoreexception.
Workaround: Ensure that client code retrieves pointers to persistent data each time it is needed, or whenever subsequent calls have been made to code that may update the persistent attribute store.
•
Problem: attribute value cache persists between runs of nbinterface
Symptom: The Northbound Event Interface's databases grow in size after registering and deregistering of many filters
Conditions: The client registers and de-registers many filters which have many objects in their scopes.
Workaround: The client can deregister all the filters, then stop Cisco EMF, delete the databases:
/db/nbinterfaceCache.db & /db/nbinterfaceOGM.db
The client will have to re-start CEMF then re-register its filters.
•
Problem: dump file in /tmp is not updated when new objects are deployed in Viewer
Symptom: When exporting an object within the domain of the Northbound Event Interface, newly created objects are not always present in the dump.
Conditions: A dump on a NEI filter is done a second time, and the objects satisfying the filter's containment scope have changed since the first dump.
Workaround: The filter needs to be de-registered and then re-registered before performing the second dump.
•
Problem: [CSCdr42275] IP Address Range Query Problem
Symptom: It is not possible to perform IP address range queries.
Conditions: Attempting to perform an IP address range query through the LocalQueryManagerInterface class.
Workaround: None
•
Problem: Query editor not updating when a map object is deleted.
Symptom: Query editor is not updated when a map object is deleted.
Conditions: When a map object that is used by an event group (the event group is collecting events on that map object) is deleted, and the Query Editor is launched on the event group, the Query Editor prompts to save the changes to the query event though the query has not been changed.
A side-effect of this is that the databased event group may retain the deleted map object.
Workaround: Re-run the query containing the deleted object.
•
Problem: AHS under heavy load
Symptom: Historical data of some objects' attributes are not being stored.
Conditions: Cisco EMF is under extremely heavy load due to performance polling, and a History Storage Criteria uses an Object Group.
New objects to be polled are added to the storage Object Group while in the order of 1000 attributes are being polled per second. The new objects' attributes may not be stored.
Workaround: After decreasing the system load temporarily, load a new History Storage Criteria, identical to the original except in name.
•
Problem: System appears to run up even tho 2 processes crashed
Symptom: A "cemf session" will not start.
Conditions: Some Cisco EMF processes failed to start. No indication of this was reported by the "cemf start" command.
Examining the "sysmgr.log" file in /logs on system startup will show if processes haven't started.
Workaround: Stop Cisco EMF. Contact Cisco TAC to find solution to problem. Restart Cisco EMF.
•
Problem: enum type in varbind shows integer value instead of text
Symptom: Alarm messages in the Event Browser have numbers in the places where strings should be.
Conditions: If a varbind for a trap is defined as an enum variable in the MIB, it is displayed as an integer value in the alarm message in the Event Browser application.
Workaround: None
•
Problem: Containment Server stores an invalid object
Symptom: It is possible for a deleted object to be added to a containment view.
Conditions: The scenario required to reproduce this is not likely to happen in deployments: it relies on an object being moved to a view at the same time as it is being deleted.
For example, using the Map Viewer an object can be copied to another location on a view using the "Add Object(s) to ..." menu option. While the user is selecting a parent destination for the object the object can be deleted using a second Map Viewer. The now invalid object is still added to the new location.
Workaround: Objects should not be moved at the same time as they are being deleted.
•
Problem: trapBuilder does not allow OID varbind values... - DEV
Symptom: The /bin/trapBuilder test utility fails to generate traps with OIDs specified as varbinds.
Conditions: A user wants to generate traps with OIDs specified as varbinds.
Workaround: If OIDs need to be specified as varbinds, a simulator should be used.
•
Problem: [CSCdr88102] Inconsistent meta data information returned by data abstract
Symptom: The list of attributes retrieved via mdsClassAttributes may differ from the attributes retrieved by getting all the attributes from all of a class's types.
Conditions: Using the the Data Abstractor to get all of a class's attributes.
Workaround: Retrieve the data by getting the attributes of the class's types, rather than by mdsClassAttributes.
•
Problem: [CSCdr88128] site objects are allowed to be deployed on network tree
Symptom: Objects of class site can be incorrectly deployed via the Corba Gateway.
Conditions: It is possible to incorrectly deploy an object of class site in the Network view through the Corba Gateway.
Workaround: CORBA clients should follow containment rules in the absence of platform validation.
•
Problem: [CSCdr90130] Two OGs with same name and in same dep. context can be dep
Symptom: It is possible to incorrectly deploy two Object Groups with the same name and put them in the same point in containment.
Conditions: This situation can only occur by using the CORBA Gateway API's to erroneously create a deployment context with two objects having the same parent and object name.
Workaround: The CORBA client should ensure that objects with the same name and parent do not appear in the same deployment context.
•
Problem: [CSCdr92254] ObjectGroup.moveObjects should move all or no objects
Symptom: When an operation to move a number of objects from one group to another fails for some but not all of the objects, the complete operation is not rolled back.
Conditions: This situation only occurs when using the CORBA Gateway Object Group API to move objects from one group to another. Note that the problem can only occur as the result of erroneous object details being entered by the CORBA client code.
Workaround: The CORBA client code needs to ensure that if it is to move objects from one group to another, all the objects exist in the source Object Group.
•
Problem: CEMF Bugs:TrapAlarmMapper does not handle Unsigned32
Symptom: Cisco EMF fails to map traps using Unsigned32 attributes.
Conditions: If a trap has attributes that are declared as Unsigned32, then it cannot be mapped to a managed device.
Workaround: Ensure that each attribute declared as an Unsigned32 is redefined to be an INTEGER in the MIB.
•
Problem: PerfPollTask intermittently create perf. object group without group name
Symptom: A performance polling object group can be created with no name.
Conditions: When a Performance polling object group is specified using a GROUP_NAME, the group is created but with an empty string for a name.
Workaround: Override the createPerformanceObjectGroup method and specify the value of m_groupName.
•
Problem: [CSCdr94148] OG.rename does not rename the OG in the OG tree
Symptom: Renaming an Object Group using the OGManagerInterface API does not rename the Object Group in the ObjectGroup view.
Conditions: If an Object Group is directly renamed via the CORBA Gateway then the name in the containment will not be automatically changed.
Workaround: If the Object Group is renamed in the Object Group then it must be renamed in containment as well, and vice versa.
•
Problem: objectFileParser fails to parse multi-line text or text with brackets
Symptom: The Cisco EMF file parser objectTypeFileParser does not support attribute value definitions spread over multiple lines nor string value definitions containing brackets, "(" and")".
Conditions: The Cisco EMF objectTypeFileParser is being run on files containing attributes spread over more than one line, or containing string-value definitions with brackets.
Workaround: Concatenate the multi-line definition onto a single line. Other brackets, e.g. <>, {}, [] can be used rather than ().
•
Problem: No progress dialog in Event B when you launch it from object.
Symptom: No query progress bar is displayed and it is not possible to cancel a query.
Conditions: This happens when the Event Browser is launched on an object from the Map Viewer.
Workaround: There is no workaround.
•
Problem: CEMF CGW IDL is not OMG compliant
Symptom: Compiling CORBA Gateway IDL files produces errors regarding symbols used.
Conditions: Compiling IDL with a strict OMG compliant compiler that produces errors regarding the overloading of names to represent different types e.g. struct, enum, enum fields.
Workaround: The workaround is to use a different compiler which is not as strict. The IDL compiles with Iona's IDL compiler for CORBA 3.0.
•
Problem: datalook get (..) rows x on mo (..) broken
Symptom: When the Cisco EMF datalook utility is used to retrieve a table attribute, and a maximum number of rows to be retrieved is specified, the entire table is retrieved.
Conditions: The Cisco EMF datalook utility is attempting to retrieve attribute details with a maximum number of rows defined.
Workaround: There is currently no workaround.
•
Problem: [CSCds06353] Possible inefficiencies in deployment process
Symptom: There is a theoretical decrease in system performance due to participant contexts being passed to participants that do not require it.
Conditions: During deployment of objects, some participants are unnecessarily called (e.g. when the participation context of the participants is registered as a ManagedObject there is no need to send this context if the level is a Provision_Creation_Tree).
Workaround: There is currently no workaround.
•
Problem: CSCds06543 CGW : Event Ohannel.- removed event type should not be sent
Symptom: When CORBA event types are removed from the channel's set of event types, the types are sent to the consumer.
Conditions: A CORBA client is registered on a channel and the eventTypes attribute of the channel is changed.
Workaround: There is currently no workaround.
•
Problem: [CSCdp54476] Renaming objects to existing ones fails without flagging ER
Symptom: The ContainmentInterface class does not return an error condition when an object rename operation fails.
Conditions: In connected views, sibling objects must have different names. Renaming an object to have the same name as one of its siblings in a connected view fails as expected. However, the ContainmentInterface class does not return an error condition.
Workaround: There is currently no workaround.
•
Problem: [CSCdr82780] Error: ObjectFileParer: create/delete failure after -upgrade
Symptom: When using "cemf remove forupgrade" to upgrade an EM, the following message can be displayed:
ObjectFileParser: create/delete failure.
Conditions: Upgrading an EM via "cemf remove forupgrade".
Workaround: This can be safely ignored. It is due to an attempt to recreate the EM's view, which already exists.
Note that if the upgraded EM's view has different properties to the original EM's view, e.g. a change to its visibility, it is the responsibility of the EM to make this change along with any other changes to the EM's object model.
•
Problem: trapManager fails with invalid trap forwarding file
Symptom: The trapManager process exits.
Conditions: Within a trap forwarding file, there isn't a space followed by a tab between the hostname and the trap pattern rule, or the headings for the rules are missing.
Workaround: Ensure the hostname and trap mapping rule in a trap forwarding file are separated by both a space and a tab. Ensure the file contains a heading of the form:
System Traps
•
Problem: Event Browser User filter on cleared alarms not working.
Symptom: Querying cleared alarms by user through the Event Browser results in some alarms being displayed that were cleared by other users.
Conditions: Different alarms have been cleared by different users. A query is invoked through the Event Browser for alarms that have been cleared by one user.
Workaround: There is currently no workaround.
•
Problem: Logger logs when shouldn't: logfiles fill with junk, making debug harder
Symptom: Logs in /logs may contain a list of the contents of ObjAttrList objects included for diagnostic purposes, marked as ERRORs.
Conditions: The ObjAttrList::dump(ObjAttrList&) method has been used.
Workaround: Calls to "ObjAttrList::dump" should be removed after EM development if these error messages are not wanted.
•
Problem: [CSCds17854] Copies of OGs created via CGW are not visible in OG tree
Symptom: An Object Group created by copying an existing one does not appear in the object group containment.
Conditions: A CORBA client calls OGManagerInterface::performGroupOperations(), with OGCopyOp specified as an operation type.
Workaround: The EM can add the group to containment via the ContainmentInterface::addObjects() method.
•
Problem: dev: Setting random data in events cause vector server problems
Symptom: The vectorServer process exits.
Conditions: A test rig raises test events on an event channel which the vectorServer process uses. The data in the events is not valid.
Workaround: There is no workaround, the scenario only occurs during development testing.
•
Problem: Alarm propagation doesn't seem to work in some cases when reparenting.
Symptom: When an object is reparented to a different view, no propagation will be seen for the alarms presently on the object. When new alarms arrive on the object, they will propagate correctly.
Conditions: Reparenting an object from one view to another.
Workaround: There is no workaround.
•
Problem: Trap Forwarding file format inflexible to white-space
Symptom: The trapManager process exits.
Conditions: Within a trap forwarding file, there isn't a space followed by a tab between the hostname and the trap pattern rule, or the headings for the rules are missing.
Workaround: Ensure the hostname and trap mapping rule in a trap forwarding file are separated by both a space and a tab. Ensure the file contains a heading of the form:
System Traps
•
Problem: ObjectStore cache manager exits with error
Symptom: The objectStore cache manager exits and leaves the following message in /var/adm/messages Cache manager ran out of space in the action set.
As a result of this client applications will crash with errors such as No handler for exception: ObjectStore internal error The Cache Manager process appears to have disappeared.
Conditions: This can occur in highly concurrent applications and may also be linked to occasions when the root partition becomes full. It occurs when the Cache Manager's fixed sized cache overflows.
Workaround: It is possible to increase the size of the Cache Manager's cache by setting the following ObjectStore environment variables. In the 'C' shell enter the following commands before running cemf shell. Alternatively, the root user can make the settings permanent by editing their .cshrc file in their home directory.
setenv OS_ACTION_SET_INITIAL_SIZE 40960 setenv OS_COMMSEG_RESERVED_SIZE 16777216
Note that the default values are 20480 and 8388608 respectively.
•
Problem: Comms ERRORs when executing many scripts or raising events
Symptom: The following messages are logged into the file 'notificationServer.log', in the /logs directory.
Gateway: WARN commsHandle.cc:259 CommsHandle::connectInternal Invalid ConnectionSpec object provided Gateway: ERROR CommsGateway::sendcommsGateway.cc:775 Failed to send message general: ERROR EventChannelManagerProxy::distributeEvents Failed to send message general: ERROREventTransferFIFO: distribute events failed general: ERROR EventTransderFIFO: failed on sending Gateway: ERROR CommsGateway::process_socket commsGate-way.cc:1247 read failed, closing connection (errstr="Connection reset by peer")
Conditions: Cisco EMF executes hundreds of scripts or events simultaneously. This is a scalability issue that depends partly on machine specification and configuration.
Workaround: Staggering notifications alleviates the symptom.
•
Problem: EventChannel Server cores when receiving events with filter setup
Symptom: The CORBA Notification Service (process called "EventChannel_svr") exits.
Conditions: The CORBA Notification Service exits when a filter is applied to a CORBA channel which supplies alarm events and the CORBA client has subscribed for this type of event.
(Alarm Events are only sent on the AlarmChangedEventChannel and AlarmAddedEventChannel)
Workaround: The client needs to remove the filter and then code a filter inside their CORBA client if required.
If a client has applied a filter and the crash has happened, the client has to restart the corbaEventChannelManger by executing...
> sysmgrClient -k corbaEventChannelManager; sysmgrClient -x corbaEventChannelManager
This does not mean that clients cannot receive alarm events, they cannot apply a filter.
•
Problem: vectorServer consumes all swappable memory
Symptom: The vectorServer may grow in size, eventually consuming all memory.
Conditions: Many alarms are raised against one object.
Workaround: Add a clear correlation rule to specify that when a new alarm is raised on an object, it automatically clears all active alarms of the same class on that object. The rule is as follows:
CLEAR_CORRELATION_RULE INCOMING_ALARM_CLASS X ALARM_CLASS_TO_CLEAR X NO_CLEAR_LAST_ALARM END_RULE
and should be saved in a file. To add this rule, run
> cemf shell > cd /bin > ./clearCorrelationFileParser
This should be performed while Cisco EMF is running.
•
Problem: object group access not updated for client
It is possible to use the Access Manager to allow certain application features to be restricted for a user. In particular, an Access Specification can be created to restrict features to objects in a particular object group.
If however, the contents of such a group changes while a particular Cisco EMF session is running, the change will not appear popup menus in, for exa mple, the Map Viewer.
This means that, where an object is removed from a group, features will still appear to be available. Where an object is added to a group, the features may appear to be unavailable.
This does not affect the actual security of the features when they are launched: at this point, object group membership is checked and access to features will be as intended.
Workaround: The workaround to correct the popup menus is to close the session and open a new one.
•
Problem: [CSCdr09383] Obj Config Dialog not moving to table and takes too long t
Workaround: There is currently no workaround.
•
Problem: [CSCdr25497] Cant view beyond 10,000 traps in Event Browser
The event browser displays a maximum of 10,000 events.
If the user has set up a query that should contain more events than this then they will see the button at the bottom left of the browser stating there are "10,000 of XXX" events, where XXX is the correct total.
The events are not necessarily the most recent events and sorting occurs on the events displayed in the browser. To see the most recent events, the user should query by time.
Workaround: The suggested workaround is to set up a query that displays less than 10,000 events.
•
Problem: [CSCdr26527] Network: object does not rename itself
Workaround: There is currently no workaround.
•
Problem: [CSCdr26799] Dialogs invoked from object list on another dialog fails.
Element Manager dialogs which have been created using Cisco EMF SDK can be launched from the popup menu in the Map Viewer. If the same dialog is then launched again, using the popup menu from the Element Manager dialog, it is possible that a different selection of objects will be displayed from those expected. This occurs when the dialog being launched from has more than one object list on it. This will cause the launched dialog to only display those objects selected in the source dialog, or any objects leading to those objects in the view.
Workaround: If this is not the desired behaviour, then subsequent dialogs should be launched in the same manner as the first, that is, from the popup menu in the Map Viewer.
•
Problem: [CSCdr16205] Network: Does not look at variable length subnet mask
The Auto Discovery server keeps track of objects and the networks they belong to using the Network view.
However, it does not make use of Variable Length Subnet Masks. Instead, it assumes that an object belongs to a network of the size implied by the class of the address.
This can mean that, for example, a class B-addressed object which belongs to a subnet will be positioned beneath a network object representing the class B network. As such, there may be many more objects underneath that network than there are in the smaller, subnet. This can make finding an object harder.
Workaround: There is no workaround to this.
•
Problem: OGM won't display path when object only in EM view
The Object Groups application allows the contents of an object group to be examined in a list within a window, with some information shown for each of the entries.
For objects which do not exist in the Physical view, the object name will not be shown.
Workaround: There is no workaround for this.
•
Problem: [CSCdp74281] NO warning if attempt to close change screen
It is possible for attributes of objects to be modified by using the Element Manager dialogs which have been designed through AVMT.
When the user modifies the values of attributes in one of these dialogs, and chooses to exit without saving, no warning is displayed.
Workaround: The workaround is to save prior to exiting the dialog, if the changes are to be kept.
This also applies to the Object Configuration dialog, where available.
•
Problem: many ERRORs seen once CEMF had been started
Workaround: There is currently no workaround.
•
Problem: [CSCdr90130] Two OGs with same name and in same dep. context can be dep
It is possible to incorrectly deploy two Object Groups with the same name and put them in the same point in containment by placing the groups in the same deployment context.
Note this situation can only occur by using the CORBA Gateway APIs in an erroneous way. The CORBA client should be fixed so that this situation doesn't occur.
Workaround: None
•
Problem: Unable to cancel any changes made in Object Group Manager.
The object group manager can be launched by clicking the "Groups" icon on the CEMF Launchpad.
When an object is created initially with no query and it is later edited by specifying a query for it and clicking the "Re-run query" button on the "object group dialog",
the object group editor does not get "busied off."
There seem to be a problem in the server code, in that the dialog does not receive the callback( the result of the re-run query operation) to indicate whether the operation was successful or not so that the dialog will be "busied off."
Workaround: There is currently no workaround.
•
Problem: (CSCat20433) AV allows TNet Radios to be re-parented to subnets
Deployment templates can arrange that different kinds of objects can only be deployed in certain places.
However, once an object has been deployed into an appropriate place, it is possible to reparent it so that it lives beneath, for example, an inappropriate container for the technology in question.
Workaround: Users must be aware of the view relationships which apply to the technology they are administering.
•
Problem: [CSCdr94148] OG.rename does not rename the OG in the OG tree
Symptom: Renaming an Object Group using the OGManagerInterface API does not rename the Object Group in the ObjectGroup view.
Conditions: If an Object Group is directly renamed via the CORBA Gateway then the name in the containment will not be automatically changed.
Workaround: If the Object Group is renamed in the Object Group then it must be renamed in containment server as well, and vice versa.
•
Problem: ObjectListViewer displays objs multiple times under certain conditions.
The Object Groups application allows the contents of an object group to be examined in a list within a window, with some information shown for each of the entries.
For objects which do not exist in the Physical view, the object name will not be shown.
Workaround: None
•
Problem: The network view (Read only) details can be re-named from any view?
Workaround: There is currently no workaround.
•
Problem: [CSCdr08600] deletion of object service fails if deployment wizard is o
If a user is in the process of deploying objects using the Deployment Wizard, for example by selecting it from the popup menu in the Map Viewer, it is not possible to delete objects at the same time.
Workaround: The workaround is to complete or cancel the deployment operation prior to performing the deletion, and then to resume the deployment if required.
•
Problem: [CSCdr22246] Image not removed from chassis view after module is delete
When EM developers define maps, no relationship checking is enforced, which means that objects can appear on incorrect maps. Objects must only appear on a map if they are immediate children of the object whose map is being defined. This is necessary to ensure that alarms do not appear twice in the same map, to avoid operator confusion. If object are added which
do not meet this criteria, then when they are deleted CEMF will not remove them from the map, because it does not expect them to be there.
Workaround: None
•
Problem: [CSCdr59277] dialog hangs if clearing event button is pressed for norma
If the user opens the Full Event Description dialog from an event on an Event Browser, and then clicks the Clearing Event button, the dialog can sometimes hang.
Workaround: This is only under particular circumstances and the user will then only be able to close the dialog via the Window System close.
•
Problem: [CSCdr69348] Event Browser does not close even if CEMF window is closed
Workaround: There is currently no workaround.
•
Problem: Deletion of object from map don't remove link object from containment
Using the Map Builder, it is possible to create a link object which joins two managed objects on the same map.
If one of these linked objects is deleted, the link object will not be automatically deleted, though it will no longer be shown in the map.
Workaround: To arrange that it is deleted, the map should manually be removed, using "Delete objects..." from the popup menu, prior to deletion of the object which is linked.
•
Problem: cemf hangs after object deletion.
In the Viewer application, the dialog can sometime hang when objects are deleted. It has been noted most commonly if the user deletes objects immediately after opening the Viewer but it is a timing issue and so may happen at other times.
Workaround: The only workaround for this is to close the Viewer from the Window system menu.
•
Problem: Event Browser sorts only on alarms present in the cache
The Event Browser can only display a maximum of 10,000 events. It is also possible to sort the events displayed in the Event Browser.
If the user attempts to display more than the maximum displayable then he could be mislead, for example, if 10,000 minor events are raised and the Event Browser displays them and then 10 critical events are raised, these will not be displayed and it may mislead the user into thinking there are only minor events, when it is simply that he is showing the maximum amount of events.
Workaround: The suggested workaround is to use the query of the Event Browser to display a manageable number of events that is less than the maximum.
•
Problem: Object Gui logs misleading messages after object deletion
When objects are deleted from Cisco EMF, is is possible that the Map Viewer application will generate error messages in its logfile, objectGui.log, under CEMFROOT/logs. This does not indicate a malfunction in the deletion operation; these messages can safely be ignored.
Workaround: None
•
Problem: Edited query not saved for query populated Object Group
When a user edits a query populated object group, unless they select to re-run the query at that point, the query will not be saved.
To edit an object group, the user will open the Object Group Editor. From here, he will launch the query editor dialog, change the query and on closing this dialog, he will be prompted as to whether he would like to save his changes. This will return him to the Object Group Editor. If the user simply closes this dialog, then the changes he made to the query will not be saved.
Workaround: The workaround is that when the user returns to the Object Group Editor after editing the query, he should select the menu to re-run the query at that point.
•
Problem: Remove object service is sometimes not displayed
Workaround: There is currently no workaround.
•
Problem: Access manager Read Only service still allows User to edit maps?
Workaround: There is currently no workaround.
•
Problem: Insufficient colors dialog on session start is hidden.
If a Cisco EMF session is opened on a workstation where there are insufficient colors available, for example as a result of another application using many already, a warning dialog is displayed.
However, sometimes this dialog can be obscured by the Cisco EMF Launchpad window.
Workaround: The workaround is to move the Cisco EMF Launchpad away to one side, allowing the warning dialog to be dismissed. Instructions for moving windows can be obtained from the documentation for the window manager being used.
•
Problem: 1st time Close button is clicked on Event dialog and event browser cores
Workaround: There is currently no workaround.
•
Problem: Event group icon and event browser not auto populating when filter chan
Editing an Event Group in the Event Groups Applications available from launching 'Event Grps' from the Cisco EMF launch Pad )
When the user edits the Event Group query, the Event Group is not automatically updated with the events that match the new query. The Re-run query menu option needs to be invoked. Similarly if there is an Event Browser open for the Event Group this will not update until the Re-run query menu is invoked.
Workaround: The workaround is to use the Re-run query to ensure the Event Group is up-to-date.
•
Problem: Event group can be created with invalid characters
Creating an Event Group in the Event Groups Application available from launching 'Event Grps' on Cisco EMF launch Pad)
When you create an Event Group and the name has invalid characters, the user is not informed that there was a problem but the Event Group does not appear to be created, i.e. the user will not be able to see it on screen.
However, the Event group has been created and so if the user attempts to create an Event Group with the same name, then they will be informed that an Event Group with that name already exists.
Workaround: The workaround is not to attempt to create an Event Group with an invalid name.
•
Problem: Edited event group name can contain invalid characters
Editing an Event Group in the Event Groups Application available by selecting 'Event Grps' from the Cisco EMF Launch Pad )
When you edit an Event Group and enter invalid characters for its name, then the tree display on the left hand side of the dialog shows the original name, but when you open an Event Group Editor you see the invalid name.
Workaround: The workaround is not to attempt to give the Event Group a name with invalid characters.
•
Problem: Unexpected behavior when trigger cond = reset cond
The Thresholding Regime Application (available from launching 'Thresholds' from the Cisco EMF Launch Pad )
If the user has set up a thresholding regime with a threshold that has a trigger condition that is equal to the reset condition and is checking this frequently, then it is possible that two notification profiles can start running when there should only be one. This may result in being informed twice about the same threshold condition.
Workaround: The workaround is not to set up a threshold where the trigger and reset conditions are equal.
•
Problem: Event group editor takes longer to close when edit query
Editing an Event Group in the Event Groups Application ( available from launching 'Event Grps' from the Cisco EMF Launch Pad). When a user edits the Event Group in any way, it can take up to 6 seconds for the Event Group Editor to close.
Workaround: The workaround is to be aware that it can take this amount of time for this opera
Guest
