The latest version of NMIS can be downloaded from the Opmantek.com Download Page
Table of Contents |
---|
NMIS 8.6.0G
This version is scheduled for release in Mid-October 2016.
Highlights for the 8.6.0G release
8.6.0G is a major release with lots of new features, improvements and bug fixes.
- NMIS now supports collection of data from Windows systems using WMI (Windows Management Instrumentation).
WMI can be used independent of or together with SNMP, and the Windows-related models have been updated to make use of both sources. - NMIS now supports newer versions of the common OS distributions better, and the installer also handles systems with a minimal base (i.e. without Perl).
- NMIS can now be installed or upgraded in a fully automated fashion without any user interaction.
- ConnectWise's REST API was added as new authentication method.
- Users can be locked out after a configurable number of failed login attempts.
- Service testing was made more flexible and can now be performed at custom frequencies.
- There are several new features for improved scalability and performance, especially for devices with lots of interfaces.
- Lots of new or improved device and device capability models, and support for inventory collection on certain types of devices.
- The standard models have been updated for greater modularity and ease of modification, and there are some flexibility enhancements for device modelling as well.
- And of course lots of bug fixes and improvements, the most important of which are outlined below.
Installation and Integration
The installer now asks only once for package installation confirmation, not repeatedly.
Better support for recent operating system distributions
This now includes Debian 6 and up, Ubuntu 10 and up, CentOS/RedHat 6 and 7.- Better bootstrapping for systems without Perl installed
As nmis depends on perl, the installer offers to bootstrap a standard Perl installation in that situation.
Now supports non-interactive installation and upgrade
To automatically answer all confirmation dialogs with the default response, pass the installer the option-y
, i.e.sh nmis8.6.0g.run -- -y
.The supported migration operations were extended and reworked for greater robustness and safety
These now include updating of the RRD directory layout, Table configuration files, NMIS plugins, and all the model and graph files.Conversion to JSON db files is now offered on upgrade
This is highly recommended for performance reasons; for initial installations it's on by default.example cron snippet files are now saved in install
The installer can now replace an existing installation if desired.
After asking about creating a backup of your existing installation you're prompted with the choice of upgrading (default) or total replacement.
GUI Improvements
NMIS can now use ConnectWise's REST API as authentication service.
Support for locking out users after N failed logins
This is controlled by two new configuration settings:auth_lockout_after
(default: 0, ie. no lockout) sets the number of acceptable consecutive login failures, andserver_admin
(a comma-separated list of email addresses) sets the administrators who are notified by email if a lockout occurs.
To clear a lockout, an administrator has to select the "reset login count" option in the User editing GUI (System Configuration -> System -> Users).Collect and update times are now collected and graphed for every node
These can be accessed for an individual node by selecting type "polltime" on the Graph Drill page. An overview is als available in form of the new report type "Collect/Update Time" (both as on-demand/Current and Saved/Historic report).- Nodes can now be renamed from the GUI
This was already possible using thenode_admin
cli tool; now you can make that change on the Node Editing page.
All files related to the node are renamed as needed. node name wrapping (for long node names) is now on by default.
Colors for the group status displays were updated
In coarse mode node downs are now shown red. The color graduation was adjusted for better scalability.- noncollecting interfaces are now easier to spot
The 'all interfaces' view now displays interfaces with collect=false with a grey background.
The service status display now indicates the service test age with color, and has more flexible display and navigation options.
A service that hasn't been checked for longer than 1.5 times its configured service frequency is shown in yellow, or in red if the age is above 2 times the desired frequency.
There are now links for easier navigation between all service-related screens, and the service selection can be narrowed down to only working or only services with problems.The find tool was improved for better flexibility, the log viewer and many oder widgets were made more robust
The find tool now provides links to the nodes or interfaces found.The NMIS Selftest now includes checking all file and directory permissions
Invalid file permissions are a relatively common problem. The selftest checks permissions only once every two hours for efficiency reasons, so you may want to clear the selftest status cache after making permission adjustments (i.e.rm /usr/local/nmis8/var/nmis_system/selftest.json
).New support for separate "display name" properties for both nodes and interfaces
In the Node Configuration GUI you can now optionally set a "display name" for a node and each of its interfaces.
If such properties are set they are used instead of the node or interface name when showing the thing in the GUI.Service-only nodes now have a better default view
The node display screen now shows more uptodate data
In the past certain entries were only updated when a type=update was performed, not on collect. This has been changed.events can now be closed easily from a node's view
If the user has admin rights, a small close icon is shown next to each open event for a node. Events closed thus are logged as 'closed from GUI'.The Network Metrics and Health widget now supports a group limit cutoff
In environments with huge numbers of group this widget could consume excessive amounts of cpu time. The new optionnetwork_summary_maxgroups
(default: 30) controls how many groups are shown by default. If the limit is exceeded, a link for an unlimited view is shown.- NMIS no longer honors the deprecated option
auth_require=false
.
Engine
Model file caching was added for improved performance
If you are making model changes you will have to run a type=update operation to refresh that cache.nmis type=update has new
force=true
option
If set, all existing node information (except RRD databases) is ignored and the node status is reinitialized completely from scratch.The interface speed limit enforcement is now configurable
NMIS has long supported to ignore invalid or nonsensical interface counters, but in version 8.6.0 this is an automatic standard feature.
Every nmis type=update operation now automatically enforces all (new or modified) interface limits, by updating the RRD file limits according to the selected enforcement policy.
The Node Configuration GUI lets you set the desired enforcement limits: optionoff
disables limits for the given interface, optionnormal
(which is the default) sets the limit to 2 times the configured interface speed, while optionstrict
sets the limit to be precisely the configured interface speed.All Down events now include outage information for easier external consumption
The event details will include text similar to this example:outage_current=true change='comment from the outage setup'
SNMP Version 3 support was improved, and SNMP V3 Contexts are now supported.
Role types, Node types and Network types are now fully customizable
In the past there were two or three hardcoded choices for each of these, and NMIS' internal logic depended on particular values.
This is no longer the case. You can use the Configuration GUI to edit, replace or add new values for each of these properties.Better integration of plugin code and main SNMP functions
All plugins now use the improved standard SNMP code, which makes extending plugins much simpler.Better control over retaining of historic,closed events
If the new optionkeep_event_history
is set to false then NMIS cleans up old closed events immediately. Default is true, i.e. old events are kept for archival purposes until the purging policy causes their cleanup.Improved purging of old files
The nmis file cleanup script was removed (as being too brittle and inflexible), and a new operationtype=purge
was added tonmis.pl
. This operation is much more precise and reliable, and also supports the optionsimulate=true
, where it only shows candidate files for purging without making any changes.- The content of test emails was improved.
- Improved
webtest
Service test program
Thewebtest
script now supports https better, also offers explicit proxy selection, and can optionally make its requests with caching off. In addition to that, the size of the response data is now reported to (and stored by) NMIS. - Some new Service test programs
For example NMIS now ships with a test program for testing Citrix Netscaler logins. - Service tests can now run at custom frequencies, independent of the collect cycle
(This does not apply to SNMP-based services, which are always and exclusively tested during a collect operation.)
NMIS now honors your choice of service Poll Interval configuration, and the service will be tested no more frequently than requested.
Service tests are performed as part of the collect cycle, but you may enable a more frequenttype=services
operation in the standard NMIS cron snippet.
RRD parameters are adjusted to suit the service frequency, but only on service creation. - SNMP-based service tests now offer improved precision and flexibility
It is now possible to check particular process instances, based on their command line arguments. Besides that, all instances of a given process are considered (instead of just the last listed process as in the past). reporting of the interface availability of a down interface is now configurable
The new configuration optioninterface_availability_value_when_down
can be used to store the value 0 instead of the default value "U(ndefined)". This applies to interfaces that are down, and affects RRD and the graphing: RRD ignores periods where the value was "U", doesn't draw anything for such periods and the overall figures (in the legend) are also automatically based on just the non-U(ndefined) periods. On an interface that is down for a long time this may lead to certain legends reading "N/A" or "NaN". If this is undesirable, the value 0 can be saved in such cases.NMIS now automatically ignores interfaces that have been down for N days or more
The new config optionglobal_nocollect_interface_down_days
(default: 30) controls whether NMIS should stop attempting to collect interfaces that have been down for more than N days.
Modelling Changes and Improvements
- New and improved device support
Amongst other this includes Alcatel OSS, ASAM, Cisco 10000 BRAS, Checkpoint, Riverbed, Fortinet, Huawei routers, Extreme, NetEngine, NovelSat, NetGear and Trango devices. - New and improved device features
VTP, MAC tables, LLDP, CDP, better Cisco environment sensor collection (fan, temperatures, PSU) - New support for inventory data collection
Specifically from Cisco and Alcatel devices - Standard models now with better distinction between interface name and interface hardware description
The GUI now shows interface properties with a suffix tag to indicate the source where applicable, e.g. "Description (ifAlias)".
Various models were amended to use the "ifName" property as unique interface name instead of the "ifDescr" whoose descriptive text is occasionally too long and unwieldy. - WMI is a new data source in NMIS 8.6.0
WMI support is documented on a separate page. - Various new alerts were added
E.g. excessive clock drift - The storage statistics collection now includes remote filesystems.
- Most standard models were overhauled for greater modularity
You will find many common modelling features are now encapsulated in their ownCommon-abc-xyz.nmis
model file, and only referenced for automatic inclusion in the individual device model files (i.e. no longer embedded).
This makes it much easier to manage changes and customizations. - new Model Policy infrastructure
In addition to greater modularity in the models, we've also added the ability to dynamically select or deselect certain model features with a flexible Model Policy.
For example, this allows the model behaviour to be adjusted for a specific node or based on other dynamic considerations.
The model policy is documented in more detail on a separate page. New Skip Interface Feature
For very large devices it can be desirable to totally ignore many if not most interfaces. The model optionsskipIfType
andskipIfDescr
control this feature. The modelModel-Cisco10000
contains a good example of this feature.- Collected data can now optionally bypass RRD storage
The new modelling optionnosave
makes a collected model variable available for use in other variable'scalculate
expressions, but it won't be saved in any RRD database. - More custom variables are now supported in dynamic model expressions
In the past mostcontrol
,calculate
andtest
expressions supported one or at most two custom variables. This has been generalized and now at least ten such variables are suppored (CVAR0
toCVAR9
). - Various graph updates
Numerous graphs were reviewed and amended with better labels, comments and the like, and the numeric precision in some graph legends was increased.
NMIS 8.5.12G
This version of NMIS was released on 8 June 2016. It fixes two security issues and contains a number of (mostly minor) improvements and bug fixes.
...
- The GUI now produces more detailed window titles and shows the configured NMIS server name in a number of views, including the main dashboard.
- The GUI now works better with limited width screens, and "systemhealth" measurements and "diagnostic" node links are now shown in a menu.
- SSH-links for nodes were added.
- A number of navigational improvements were added for handling widgets that were expanded onto their own pages.
- The self-test is now more precise and displays better data in the GUI.
- Disk usage is now shown more human-friendly.
- A new view called Node Admin Summary was added, which displays an overview of nodes from an administrative perspective (e.g. does it seem to be configured correctly? does SNMP work? are there any administrative issues with a node?). This is accessible from the menu "System" -> "Configuration Check" and on the group lists under "Network Metrics and Health".
- The main login screen now displays a list of installed or available NMIS addon modules. (If necessary this can be disabled by setting the option "
display_module_list
" under "modules
" in the main Config.nmis.)
Engine
- The handling of nodes with large numbers of interfaces was substantially improved, and a number of optimisations to reduce the collect times for such devices were implemented.
- The event subsystem was reworked for improved performance, and a number of related bugs were identified and fixed.
- NMIS now uses locks for per-node operations to ensure correct sequencing even if an update or collect operation takes a very long time.
Time-exceeded events now also include better diagnostic info, i.e. the node name in question. - Improved Threshold Alerts for some measurements, e.g. CPU.
- New and improved models for Checkpoint, APC, ACME, Furukawa, Alcatel devices, Ubquiti, Microtek, Frogfoot, Cisco; better modelling of OSPF.
- A number of models have been reworked for greater modularity and ease of use. Many were adjusted to handle corner cases better.
- Updates of the Cisco product mibs.
- For newly added nodes a type=update operation is now run automatically if and when required.
- NMIS now detects problems with sending events via syslog better and logs them.
- Node Configuration overrides are now presented better in the GUI, and the nodeconf subsystem was reworked for better performance.
...