The following defects have been fixed since version 6.0 was released.
NT only fixes:
Solaris only fixes:
HP-UX 10.x, HP-UX 11.x, Solaris only fixes (no NT):
Symptoms: NNM_00159: Cumulative Consolidated Patch
NNM_00155:
On Japanese NT systems, the default menu for an xnmbuilder
application was
"{Japanese characters for "Configuration"}->".
If you used this menu for an application, it resulted in an
extra "Configuration" menu on ovw, the only difference being
was that it didn't have the mnemonic "(C)". This was not a
problem on Unix systems.
NNM_00154: Errors are reported back by the ovEvent stack ( pmd ), ovactiond, and other programs when a trap uses FORWARD and the file specified is empty. The man page says that this is allowed, and there are cases, such as installation, where this is needed. This fix is to allow for empty files to be used.
NNM_00146: snmpColDump -tTI did not return years as 4 digits. This patch forces years to 4 digits. Customers who use scripts that rely on 2 digit years should change their scripts. This is for Y2K compliance.
NNM_00145: If the ECS config gui is brought up directly from the URL, http://hostname/OvDocs/C/ecs/ecscmg.html, the login page is bypassed if security is turned on. Even if security is turned off, the ECS config gui will come up when the session manager is not running. This is in error.
NNM_00144: Double-clicking on a symbol in an existing submap will end up selecting any underlying symbol in the child submap that gets loaded as a result of the double-click.
NNM_00143: During startup of the _debuggable_ version of jovw.jar, an "IllegalAccessError" is thrown and a stack trace is printed to the Java console in the browser. Network Presenter "Initializing..." progress window remains at 40%. Network Presenter never comes up.
NNM_00142: Network Presenter does not allow navigation to submaps associated with connections or meta-connections. Meta-connections are not represented in scope pane. Network Presenter does not sort items in scope pane or tabular view. '+' indicator does not disappear after moving the location cursor (via click or keyboard navigation) to a non-expandable entry (one with no submap or containing no children) in the scope pane. Location cursor in scope pane can be disassociated from actual submap represented in content pane. Navigation in scope pane is inadequate and incorrect.
NNM_00141: Network Presenter map does not represent line types and special graphics found in ovw. Dashed/dotted lines and those with graphics (fiber links, tcarriers) will show up as plain solid lines.
NNM_00140: On Japanese Unix systems, the Fault->Ping, Fault->Network Connectivity->Poll Node, Fault->Remote Ping dialog boxes were too wide (filled the entire width of the display.) This was not a problem if LANG=C or on NT.
NNM_00138: When hiding a connection symbol (metaconnection) without a symbol label ovw would core dump on solaris.
NNM_00137: Missing a data point in the Collector db every time any agent's counter wraps.. snmpCollect assumes that the agent was reset, when that is only 1 of the possibilities. Other case we should be looking for is when the counter has reached the maximum and he wraps it. This patch tries to make a reasonable guess as to whether it wrapped (in which case calculate the counter assuming it wrapped), or whether it was reset (in which case blow away one data point as we did before).
Info from customer: They had FDDI cards and were collecting on ifInOctets, which were racking up at around 2.5MB/sec. At this rate, the old 32-bit SNMPv1 counter wraps in 30 min! Note the newer SNMPv2 64-bit counters will alleviate this.
NNM_00135: Customer could not enter ip addresses or ip wildcards as an event source.
NNM_00134: Given certain operations such as save and restore of the state file, ovalarmsrv would grow in size. This is due to a string not being freed along with its parent structure.
NNM_00133: 1. Certain network devices (have seen some from Cabletron and 3Com) can never get discovered.
2. When setting the status polling interval to an extremely large value (~ 3000 weeks), netmon would flood interfaces with status polls.
3. If you turn on snmp tracing, you will see netmon doing level 2 interface status polls on unmanaged interfaces.
4. If NNM is managing a Dell MSCS cluster (High availability cluster), and the cluster IP address migrates from one node to another within the cluster, sometimes NNM will rename the old clusternode to have the name of the cluster IP address, and the node currently supporting the cluster IP will not show it on the map.
5. Selecting node on map and pulling up web browser on management URL may give not-found error.
6. Netmon process may dump core in HSRP environment.
7. Non-IP interfaces are not status-polled when node is newly discovered.
8. NNM does not reschedule topology poll for switch device that generates a spanning tree protocol trap.
9. There is no way to tell NNM to not discover L2Nets.
10.There is no way to tell NNM to not schedule SNMP polls for status checks on non-IP interfaces.
11.When IP address comes up on certain nodes, a configration poll is supposed to be scheduled, but isn't.
12.There is a lot of console output when using ECS Router Down.
NNM_00132: Generally this is a defect seen only by devlopers, as it would be caught early on in the development phase. Symptoms will vary depending on submap type used.
SegSub::queryAddNode: getSegmentById() failed
The above error was seen by a developer in nettl.LOG00. This occured when he tried to add a node to his submap. We see the error was in a method of the segment submap class. The developer was using a submap type of 4 which to ipmap is a segment submap. Since ipmap did not create this submap initially it should not be of type segment submap
IPMap can take a (relatively) long time to synchronize maps with network devices with lots of interfaces.
NNM_00116: When binary datafiles are recreated using reading of data from stdin, the IP addresses get reversed on NT. This is due to the behavior of the inet_addr function. An example is that 15.2.117.131 becomes 131.117.2.15 in the file.
When the following sequence is performed, xnmgraph would give an exception. 1.Data collection setting: store, no threshold instance: all. polling interval: 10s 2. peformance->graph SNMP data -> all 3. then, data renewal by clicking on "Update Data" menu. If user keeps repeating the 3-rd step, grapher would crash.
The Object Properties/Attributes dialog box now sorts the attributes for both Name and Value. Clicking on the "Name" header sorts/ reverse sorts the attributes by name, and clicking on the "Value" header sorts/reverse sorts the attributes by their values.
If you used the panner on a submap, then closed the panner and opened a new window, on occasion, there would be garbage/ transparent regions where scroll bars might go if they were needed. This problem only happened on NT.
xnmsnmpconf wrongly displays time indications, and can update wrongly. When the user enters a time for "default polling" ( eg ), in units less than those it could be displayed in ( eg 28 hours ~ 1 day, or 10 days ~ 1 week ), xnmsnmpconf when redisplaying the data will display it in the larger units, but as an integer number. This is fine, PROVIDING, no other data field is updated and the form will be saved to the file, overwriting the original 28 hours or 10 days.
In NNM/NT, ovw can't run a registered application if a directory in the command path contains spaces, for example "Program Files".
trapd.log on NT has occasional entries looking like: ... [1] private.enterprises (IpAddress): 177.113.2.15 ... The IP address should appear as "15.2.113.117" instead. This happens whenever an incoming trap contains a VARBIND of type "IPADDRESS". ... [1] private.enterprises (IpAddress): 177.113.2.15 ... The IP address should appear as "15.2.113.117" instead. This happens whenever an incoming trap contains a VARBIND of type "IPADDRESS".
On Japanese systems, the dates for events in the xnmevents browser were wrong, sometimes having extra parentheses as well as missing the characters for month (gatsu) and/or day (nichi).
Upon startup of ovw, some submaps with autolayout turned off would have their symbols re-laid out anyway
Synchronization failure in ovspmd's SIGCHLD handler sometimes causes updates to ovspmd's tables to be interrupted.
NNM_00115: Ovfiltertest will coredump or get an out-of-memory condition when testing filter with certain size databases. There is no relation to the filter being used. This is due to topology objects loaded not being freed.
OV_BACKGROUNS_PATH should have been OV_BACKGROUNDS_PATH.
If NNM is managing a Dell MSCS cluster (High availability cluster), and the cluster IP address migrates from one node to another within the cluster, sometimes NNM will rename the old clusternode to have the name of the cluster IP address, and the node currently supporting the cluster IP will not show it on the map.
In using the "Set Filters" dialog, the buttons are not disabled during processing of buttons already depressed. Given this, it is possible to queue up the buttons ( multiple X events ). The code did not account for this and it tried to use a structure that was freed by the processing of a prior button being depressed.
snmpwalk of remote console will return snmp information from the server. If the management station is x, and the remote console is y, then running snmpwalk y system on y will return snmp information from system x. This is due to invalid value returned from function to see if localhost should be used.
If an application used OVwRemovePopupMenuItemFunction() to remove a menu item from the popup menu, ovw would occasionally core dump upon exit or opening a different map.
When setting the status polling interval to an extremely large value (~ 3000 weeks), netmon would flood interfaces with status polls.
User selects unmanaged node on map which has multiple interfaces, and manages that node. Netmon may later unmanage some of the interfaces when it should not. Mostly happens on Japanese NNM.
Defect Description: NNM_00159: Cumulative Consolidated Patch
Resolution: Cumulative Consolidated Patch
NNM_00155:
The default menu should have been
"{Japanese characters for "Configuration"}(C)->"
NNM_00154: The defect is that the man page and the behavior of the trapd.conf reading routines are different. It appears that there are justifiable reasons for having empty FORWARD files, so the code should have allowed them.
HPUX 11.X changed the format of the message header for the size of the uid field. This causes a coredump due to improperly processing the data segment of the message by netfmt.
ovdwevent -trim did not use the default values for deleting the detail records and aggregate records.
ovdwquery -sep causes coredump. This is due to printing out the the character was being done using %s instead of %c.
remote consoles are not supposed to be using ov.conf file for address resolution.
nmdemandpoll on remote console should have not have used the ov.conf file.
snmpColDump should have printed date using 4-digit year.
Map customization will stop if unknown symbol type is found. This should have been a warning and the code should have continued.
ovfiltertest would coredump due to running out of memory. This was due to not freeing topology objects after using them.
Allow empty forward files for trapd.conf. This is needed for applications that add traps with the FORWARD parameter specified.
NNM_00146: strftime with the %x option returns mm/dd/yy for LANG=C. This needs to be mm/dd/yyyy for Y2K compliance.
NNM_00145: None
NNM_00144: Event handler was allowing further processing of the event after the double-click.
NNM_00143: The problem is due to a data member being declared as private and then is accessed by an inner class (should be legal, but the Java VM in Internet Explorer and Netscape Communicator do not like it).
NNM_00142: Features were simply not implemented in original code.
NNM_00141: Features were simply not implemented in original code.
NNM_00140: One of the input fields did not take into consideration multibyte characters.
NNM_00138: ovw assumed there would always be a connection label so when it did a strcasecompare on solaris with a NULL entry it core dumped.
NNM_00137: None
NNM_00135: Customer could not enter ip addresses or ip wildcards as an event source. Customer can now use ip addresses as event sources. Wildcards will also work, including the following types:
15.*.*.* 15.1-255.1-255.1-255
NNM_00134: None
NNM_00133: 1. If the device reports its only IP address as associated with an interface whose ifType is softwareLoopback (value 24), then that interface will not be connected to any network. Since it is the only IP address, NNM can not handle a device that is not connected anywhere, and the discovery fails. This patch provides a new oid_to_type flag that can be used to indicate that nodes of this type need to have their software loopback interfaces treated as if they were a normal interface, and hence are left connected to the network. The new flag is the L flag.
2. The status polling interval was not being checked, when calculating with this value MAXINT could be exceeded resulting in a negative time. This caused netmon to ping the machine continuosly.
3. Netmon was using the node status to decide whether or not to do level 2 interface polls.
4. NNM is not reacting nicely to the fact that the Microsoft SNMP agent has an entry in the ip.ipAddrTable for IP address 0.0.0.0 on MSCS cluster nodes. This change will cause NNM to ignore such entries in the ip.ipAddrTable.
5. Netmon was adding an extra / to the end of the management URL stored in the ovwdb.
6. There is a timing condition where netmon could discover an HSRP interface on a Cisco router, which would cause the router to be deleted and re-added, and then cause netmon to abort with a core dump.
8. Spanning tree traps were not making it into the netmon process.
9. The discoverLevel2Nets command line option has been implemented for netmon. See the netmon man page.
10.The nonIPStatusPolls command line option has been implemented for netmon. See the netmon man page.
11.Netmon was making a wrong decision regarding the re- scheduling of the configuration poll for a node when one of its interfaces came up.
12.There were a lot of debug statements left in the ECS Router Down code.
NNM_00132: Upon synchronization ipmap would try to create submaps of type submapType regardless of whether ipmap was the original creator of the submap.
If a node with multiple interfaces had one interface change while the map was closed, it would appear as if all the interfaces had changed.
NNM_00116: The "Data Update" option when called does an opendir() and readdir() repeatedly upto a maximum 'maxlines' times. opendir() in the portability layer was setting a limit to the MAXDIR that can be opened to 64. This is too less when there are a large number of data files in the SNMPCOLLECTDIR.
The functions for determining the window size were getting called with the wrong values.
The Poll interval is an integer so the conversion to the larger units is wrong. Also whenever any field in the GUI is modified, a function is wrongly called which converts the new displayed larger units into seconds and stores the values in the file. So the original value is lost.
It is not possible to quote paths within the command definition. Code has been changed so that paths quoted as follows work correctly: Command -initial "\"C:/test test/notepad.exe\"";
xnmevents was not using the correct format for the date for each three platforms.
a map customization feature was forcing the autolayout of symbols when autolayout was specifically off.
NNM_00115: NNM is not reacting nicely to the fact that the Microsoft SNMP agent has an entry in the ip.ipAddrTable for IP address 0.0.0.0 on MSCS cluster nodes. THis change will cause NNM to ignore such entries in the ip.ipAddrTable.
The status polling interval was not being checked, when calculating with this value MAXINT could be exceeded resulting in a negative time. This caused netmon to ping the machine continuosly.
The exit code that cleaned up the menus was causing a NULL pointer dereference
Netmon is not properly clearing an internal flag, which causes it to update information in the topology database incorrectly, including marking a managed interface as unmanaged.
Symptoms:
PHSS_16454:
Generally this is a defect seen only by devlopers, as it
would be caught early on in the development phase.
Symptoms will vary depending on submap type used.
SegSub::queryAddNode: getSegmentById() failed
The above error was seen by a developer in nettl.LOG00.
This occured when he tried to add a node to his submap. We
see the error was in a method of the segment submap class.
The developer was using a submap type of 4 which to ipmap
is a segment submap. Since ipmap did not create this
submap initially it should not be of type segment submap.
Ipmap will allow symbols for remote versions of objects
become either unmanaged or managed while the objects in the
topology do not change.
1. Selecting node on map and pulling up web browser on
management URL may give not-found error.
2. Netmon process may dump core in HSRP environment.
3. Non-IP interfaces are not status-polled when node is
newly discovered.
If a submap property (e.g. context) changed, any application
listening for ovwSubmapChange would get two callbacks.
ovrepld is limited to the soft limit number of file
descriptors
If you turn on snmp tracing, you will see netmon doing
level 2 interface status polls on unmanaged interfaces.
When a secondary version of a node becomes the primary
version, the status of the node may not be set properly.
With LANG=ja_JP.SJIS, the Event Object Identifier editable
field was missing from the Event Configurator dialog box.
Dragging symbols of certain types caused ovw to core dump or
use a generic drag cursor. The symbol type in question had
a
bitmap file named "root" that is the same as the "root"
toolbar button bitmap name. The symbol also had the toolbar
bitmap instead of its own bitmap.
1. If a submap's background graphic file is an invalid
pathname with no "/" in it, ovw will dump core when
you try to open that submap.
2. If a symbol has an invalid symbol type, ovw will
dump core when you try to display its popup menu.
3. Can't paste text into ovw dialog text fields (intro-
duced in patches PHSS_15927/PHSS_15926/PSOV_02061)
The Locate->Objects->By Selection Name... dialog box
would grow in width every time you selected the "Apply"
button. This especially happened with objects who had
more than three entries in the list.
The Describe/Modify dialog reads in all the object's field
values when it comes up. When the user OKs the dialog, it
writes out all the object's field values, whether any have
changed or not. It should only write out values that have
either been changed by the user or an enrolled application,
or on which the modified flag has been set.
Any ovw menu that invokes xnmappmon will show the problem
when the xnmappmon File->Save menu is selected. When the
operation is done, the file saved is always zero-length.
This started happening with the latest Consolidated Patch.
Customer had a huge /etc/services file, and it was located
on a different system. Any app using SNMP (for example,
snmpwalk, snmpget, etc.) was taking an extra 5 or 6 seconds
each time it started up. This is due to the SNMP library
always checking /etc/services, even when there are shortcuts
it can use to avoid taking this performance hit.
Several NNM modules (usually xnmevents is first) complain
about a corrupt trapd.log. Examining the log will show that
at least one entry has been split into 2 lines just before
the semicolon (all trapd.log entries must be just one long
line .... there is always a semicolon separating the last
3 fields in the line from all those before). This happens
because a newline character was not properly filtered out
of the message that appeared just before the semicolon.
The only time this problem has been seen is when pmd is
running and the trapd.conf file is moved/removed and then
"xnmevents -event" is run while the file is missing. Then
try (re-)starting the xnmevents window.
(SOLARIS) snmpCollect coredumps in libovutil due to a NULL
varbind for sysUpTime in an incoming PDU from some PC with
a bad agent.
User modified a vers. of $OV_CONTRIB/NNM/setStatus/setStatus
to add a VARBIND to override the Category to "LOGONLY" :
.1.3.6.1.4.1.11.2.17.2.6.0 octetstring "LOGONLY"
This should have caused the event to not appear in the Event
Browser (xnmevents), but it still did.
Reported on NNM4.01, but patches needed for 4.11 & 5.01 too.
xnmtrap cores on SOLARIS w/ blank format string.
It also corrupts trapd.conf as it cores.
To reproduce problem:
-Warning!! : before testing, backup $OV_CONF/C/trapd.conf
-Run xnmtrap on SOLARIS 4.11
-Click on any Enterprise Name
-Doubleclick on any Event Name under that
-Blank out the Event Log Message line
-Set Category to Don't log or display
-OK
-File->Save .... this will cause the coredump.
On Solaris, after starting approximately 60 ovw sessions
or 255 ovwdb connections (via an application), subsequent
connections are refused.
Certain network devices (have seen some from Cabletron and
3Com) can never get discovered unless the I flag is set for
them in the oid_to_type file.
This has different symptoms on different systems. In
Solaris, this may cause a coredump or the message "Cannot
connect to ovtopmd". On HPUX, this can cause the message
"Cannot connect to ovtopmd" or problems updating status of
related objects ( e.g. a set of interfaces for a particular
node ).
The program mibtable run against a non-existent nodename
will coredump. This is due to not validating that a snmp
session was actually created.
Certain 3rd party applications stop working after applying a
patch for the large value_info.pag file.
Defect Description:
PHSS_16454:
ipmap upon synchronization would try to create submaps of
type submapType regardless of whether ipmap was the
original creator of the submap.
Ipmap was allowing the managed state of map objects change
when there is no way to propagate the information to the
remote collection station.
1. Netmon was adding an extra / to the end of the management
URL stored in the ovwdb.
2. There is a timing condition where netmon could discover
an HSRP interface on a Cisco router, which would cause
the router to be deleted and re-added, and then cause
netmon to abort with a core dump.
The background graphics code sent out an extra event even
if the submap background was not changed.
No steps were taken to increase the number of file
descriptors available beyond the soft limit set by the
kernel.
Netmon was using the node status to decide whether or not
to do level 2 interface polls.
The counts of up and down interfaces were not being reset
properly.
This was caused to an incorrect buffer length.
ovw was using the toolbar bitmap file instead of the symbol
type's bitmap files, even though the symbol type had all the
appropriate bitmap files in the $OV_BITMAPS/$LANG directory.
As a result, ovw couldn't properly compute the drag cursor
for the symbol and core dumped.
1. ovw dereferences null pointer in this case.
2. ovw dereferences null pointer in this case.
3. Disabling drag (button 2) on text widgets also
disables paste (button 2).
The dialog box inadvertently resized when the list
was updated.
ovw wasn't checking whether the fields it writes out had
been modified.
The underlying communications code for NNM uses the fdopen()
system call. On Solaris, this call fails on file
descriptors greater than 255.
If the device reports its only IP address as associated with
an interface whose ifType is softwareLoopback (value 24),
then that interface will not be connected to any network.
Since it is the only IP address, NNM can not handle a device
that is not connected anywhere, and the discovery fails.
This patch provides a new oid_to_type flag that can be used
to indicate that nodes of this type need to have their
software loopback interfaces treated as if they were a
normal interface, and hence are left connected to the
network.
Certain previous patches caused the "TopM Interface List"
field values to be removed from ovwdb, thus allowing it to
run with a smaller value_info.pag file. Some applications
depend upon this field being in the database. This patch
will change the default behavior of ovtopmd such that the
field will be restored in ovwdb. If the customer requires
the field to be removed to save database space, and does not
depend upon the failing applications, then the behavior may
be reinstated by adding the -U option to ovtopmd in
ovtopmd.lrf.
Symptoms:
PSOV_02166:
When tracing in ipmap is used, it will core on Solaris if
certain submaps are opened. This is due to the formatting
of submap information not checking for app_name (
meta-connections ) and background graphics not having
values.
Defects:
PSOV_02166:
None