Menu

Show posts

This section allows you to view all posts made by this member. Note that you can only see posts made in areas you currently have access to.

Show posts Menu

Topics - Tursiops

#1
Hi,

On Windows, I am at times seeing an issue where the NetXMS service will not restart (either after an upgrade or a manual restart attempt to load a new configuration).
Checking the logs, I this is generally due to a the NetXMS agent port still being in use by PowerShell (which we use extensively for Parameters, Actions and ParametersProviders). The actual NetXMS agent service is stopped, the process is not running, but a PowerShell process which was started by NetXMS is still active and for some reason managed to hold the agent's network port open. The system does not recover from this by itself, we have to go in manually and kill powershell.exe, after which the NetXMS service will start again.
Not sure if this is an issue where the NetXMS agent either does not clean up all external commands (properly) prior to shutdown or if at some point it just loses track of one of those commands, which is then never killed and just keeps sitting there.

We've had this issue for a while and this is appears to impact maybe a handful of systems a week (I have two systems which seem to have this problem pretty much every week, other systems appear to be more random). Considering we have 2000+ agents, it's certainly not a huge issue, but it is quite a nuisance to pick up on this  (with some systems which we do not expect to be on all the time, this problem may be ongoing and we wouldn't even notice) and resolve it.

I am pretty sure we have this problem prior to starting to implement ParametersProviders and we are not regularly running actions against the two recurring systems, so I have to assume it's just an external parameter problem.

Has anyone else encountered a similar problem?

Cheers
#2
Hi,

I have a few Performance Counter DCIs, which will only collect data when a process is running.
When the process doesn't run, NetXMS will generate a collection error.
So far, so expected.

In this particular case, I'd like to be able to transform a data collection error into a value, i.e. 0. Main reason for this are graphs - the errors mean NetXMS will try to connect the dots between whenever instances when the process was running, which looks rather confusing and 0 is in fact accurate in this case. If a process is not running at all, then it is not utilising any CPU (or disk/memory).

But it looks as if a collection error means the transformation script is not even called (I added some trace commands, but nothing was ever logged)?
Not sure if it would be possible to allow sending such an error condition through to a transformation script to handle (without breaking the current process, i.e. if a transformation script doesn't handle this particular condition, it should still return the usual data collection error).

Cheers
#3
General Support / DCIs and data types
January 10, 2018, 03:52:23 AM
Hi,

There have been some changes around data types recently and it appears that the data type in a DCI is now both, the expected data type for in- and output.
For some transformation scripts that may not be desirable though. For example data might come in as a string, but the transformation script returns an integer. This integer will now be treated as a string and thus line graphs don't work (just showing "0").

Not sure if it would be possible to allow users to optionally have different in- and output data types?

Cheers
#4
General Support / Interface Oddity
January 09, 2018, 03:45:28 AM
Hi,

We have run into an odd issue after installing firmware upgrades on a number of virtual firewalls.
The old firmware would happily show all relevant interfaces, while the new firmware results in a lot of nonsense (and breaks our instance discovery).

Initially I was going to consider this a bug in the firmware and report it to the vendor, however, upon going through an actual SNMP walk on two devices (one with, one without the upgrade), both return sensible data. So now it seems like NetXMS is having an issue of some sort.
I can confirm that the firmware does lead to changes in the interface tables, namely the interfaces suddenly start at 65536, instead of 1.
For all I recall OIDs fields are allowed to be up to 2^32-1, so this shouldn't be an issue?

See attached screenshots of a device with the old and a device with new firmware.

Thanks
#5
Hi,

I stumbled over some syslog entries in our system which were being generated just before midnight on 31.12.2017 and ended up with a timestamp of 31.12.2018 in the NetXMS database. Screenshot below. The timestamp on the left is the NetXMS database one. The one on the right is part of the syslog message generated by the device.

The entries appear to all be in the one minute before midnight, so possibly a bug?

Cheers
#6
General Support / Instance Discovery Grace Period Issue
January 03, 2018, 12:18:51 AM
Hi,

I noticed that the new feature of a grace period for DCIs sets instances to disabled when they are no longer found.
However, it does not appear to re-activate them when they re-appear, which means unless one manually goes through the nodes to locate such DCIs to re-activate them, they will simply be deleted after the grace period.

Not sure if that is the intended behaviour?

Cheers
#7
General Support / Agent Upgrades
September 05, 2017, 06:52:41 AM
Hi,

It seems all attempts to upgrade agent packages via Package Manager now result in "Installing" and en eventual timeout.
When I run a configuration poll afterwards, the agents were upgraded successfully - but with the number of agents we are running, deployment is slowed down quite a bit as I have to upgrade in small batches rather than just queuing everything.

All our agents are using Agent Tunnels, so I am not sure if that is somehow related?

Cheers

Update: A single agent which was not using Agent Tunnels actually completed the upgrade properly.
#8
Feature Requests / Filtering for Line Graphs
September 05, 2017, 01:06:43 AM
Hi,

It would be a nice feature to be able to filter values on a line graph.
So instead of just setting the minimum and maximum Y values for the graph, which will still show the up/down spikes of values higher or lower than that, be able to actually set the maximum and minimum values to be returned and used for the graph. It can make a graph a lot more readable to be able to filter some data out.
For usability the system shouldn't really ask you for those values when you select DCIs and say "Line Graph", but once the graph is shown, it could be an option to go into the properties of the graph and set the minimum/maximum values per data object on the graph.

Cheers
#9
General Support / Agent Actions and quotes
September 05, 2017, 12:55:26 AM
Hi,

Do " get stripped out of parameters when calling an Action via Object Tools? I am trying to run a netsh command which requires me to place the interface name in quotes, but the command is not working and the command is logged without quotes in the agent log (debug level 7). The command also seems to be cut short, (presumably related to the quotes).

I have an action that triggers the following:
cmd /c $1
and I am passing
netsh interface ipv4 show subinterface interface="Ethernet 2"
The log then tells me it is executing this:
cmd /c netsh interface ipv4 show subinterface interface=Ethernet

And the result is that the command of course doesn't work as expected.

Thanks
#10
General Support / Logwatch - Environment Variables
August 15, 2017, 06:16:24 AM
Hi,

Just wondering, do/shoulod environment variables work in a Logwatch parser definition?
I tried using %PROGRAMDATA% and the result was that the NetXMS Agent service would not start anymore. The moment I replaced that with C:\ProgramData everything just worked. As different Windows versions use different paths, using an environment variable would be preferable to creating a number of parsers (one for each possible file location)?

Cheers
#11
General Support / SSH Subagent and Actions
August 08, 2017, 03:27:56 AM
Hi,

I know I can use SSH for DCIs. Based on the 2.1 release notes, it should also be possible to use this for Actions, but I can't find documentation on how to make this work.
Are those Actions configured in an Agent configuration file or Actions that can be configured in Object Tools or both?

Reason for asking is that I am thinking of setting up some Object Tools which will run commands on routers/switches to pull some information for troubleshooting on request (without having to SSH into the device).

Cheers
#12
General Support / Force node to not go through a proxy
August 07, 2017, 04:03:33 AM
Hi,

I have a few systems which I do want to be part of a zone, but I do not them to be polled via proxy.
The zone itself does have a default proxy set, as that makes sense for 99% of all systems in the zone.

How can I force this node to not use a proxy?
Only thing I can think of right now is to set the NetXMS server itself as "proxy"? Is that the correct way of doing it?

Cheers
#13
Feature Requests / Hyper-V Support for Topology
July 25, 2017, 07:34:23 AM
Hi,

It would be nice if Hyper-V hosts and guests would show on a Layer 2 Topology map.
The hosts obviously do, but the guests generally do not (unless they have a dedicated NIC).

I can use PowerShell to generate lists of Hyper-V guests, virtual switches and NICs and MAC addresses, but it having this whole setup visible on an automated topology map would be much more flexible.
Cheers
#14
Feature Requests / Create Node from Agent Tunnel
July 25, 2017, 07:19:00 AM
Hi,

It would be nice to have the ability to create a node from an unbound agent tunnel.
If the Agent configuration includes a Zone ID, it should be possible to search if such a node already exists in the system, e.g. by checking IPs & hostname? That could also be used to automatically bind tunnels to existing nodes if there is a 100% match.

Even just having the ability to right-click on a node in the Agent Tunnel Manager and selecting "Create Node" would be helpful when you roll out Agents to dozens/hundreds of nodes. :)

Cheers
#15
Hi,

Earlier today I created a network map and noticed that some of the configured labels on switches were incorrect, so I went ahead and reconfigured those on the switches.
I then ran an interface name poll and assumed the map would update its link labels. But it did not. The old labels remained.
I ran configuration and topology polls as well, same result.

Do link labels not get updated? Is this a one-off thing that maps do when they create a new connection between devices (i.e. check the interface names, then statically store the current value as part of the connection object itself)?

This is an automated L2 map, so I can of course recreate it, but doing so just because a link label needs an update seems excessive?
Or is that something that's updated as part of housekeeping once a day?

Cheers
#16
General Support / WatchGuard Firewalls and Topology
July 20, 2017, 02:37:43 AM
Hi,

It appears there is an issue with getting WatchGuard firewalls to show peer connectivity properly.
I noticed that checks on interfaces almost always result in the firewall supposedly being "indirectly" connected to a specific switch port - as more than one MAC appears to be behind the interface in the ARP table and the firewalls do not do LLDP or CDP.

Not sure if there is a way around that?  :-\

Cheers
#17
Hi,

I was wondering if it's possible to pass parameters to a "Server Script" called via Object Tools (and reference them inside the actual script).
I know I can pass an input field as %(inputfield), but while I can make this work with an Agent Command, I seem to have trouble with Server Script types. Has anyone done this and give me some pointers?

Thanks
#18
Feature Requests / LAGs and Peer Nodes
June 02, 2017, 08:21:23 AM
Hi,

I noticed that the Interfaces tab seems to only accept 1-to-1 links.
When two switches are connected via LAG for example, only one of the interfaces one each switch will show as being connected to the other end. Same with servers connected via LAG.

We also encountered a network the other day, which did have a network loop in it. STP had blocked one of the interfaces so there was no obvious impact.
In this case, NetXMS was showing interface A on switch A to be connected to interface X on switch B - and the interface was down.
Switch A's interface B was also connected to switch B on interface Y - and this interface was up. However, NetXMS did not show peer nodes on those interfaces.
Interface A's peer node had been detected via STP, while Interface B was showing in LLDP.

I am not sure if NetXMS' design allows for handling of n-to-n relations when it comes to peer nodes?

Cheers
#19
General Support / 2.1-RC1 Console - slowish
May 24, 2017, 10:05:36 AM
Hi,

Since the upgrade to 2.1-RC1, the NetXMS Management Console, including the web interface seems to be significantly slower than before.
Especially alert handling. Not sure if that's related to the new Zone column being added to it (and us having a lot of zones & alerts).

Has anyone else encountered/noticed this?

Cheers
#20
Feature Requests / Node "isPhone" flag
May 23, 2017, 05:29:33 AM
Hi,

Would it be possible to add an "isPhone" flag to nodes?

For phones which do not support SNMP (or where the vendor has locked it down), this could maybe be passively determined via switches CDP/LLDP data during a topology poll? I can't see the device type when I check the LLDP table on a switch in NetXMS, but on the managed switches themselves, the data appears to be there.

Cheers