NetXMS Support Forum

If there's NetXMS or SNMP agent on the end node, MAC address will be read from it. Otherwise we take it from FDB of the switch to which this node is connected.

А можете поставить в конфиге агента
DebugLevel=3

перезапустить его и прислать лог за первую минуту его работы.

Это именно на Windows системах? Какой версии Windows? Сколько ядер/процессоров?

Well, my bad, seems that I did not fully test the script that I've posted initially. It should be like this:

function main(oid, name) {
  transport = $node->createSNMPTransport();
  if (transport == null) return null;
  v = transport->getValue(oid);
  h = %{};
  for (a : v->split(",")) {
    b = a->split("=");
    if (b->size == 2) h[b[0]] = b[1];
  }
  return h[name];
}
The way it works - v->split(",") produces array where elements are "voltage=121.0" and so on. We loop through that array, so variable "a" has the value of each element of that array.
Now we need to split by "=" character and variable "b" becomes an array with two elements: [voltage, 121.0]. To access these elements we use index, so b[0] is "voltage" and b[1] is "121.0".
Now, h is a hashmap - sort of an array, but instead of numeric index elements are identified by string. We initialize it empty with h = %{}; And then we fill h with values with this code: h[b[0]] = b[1]; (and I now see the reason of the wrong code - forum editor is garbling the h[b[0]] part when pasting)

I'd guess the unit is ms, but I'll check.

And if your try policy as simple as that:
<parser checkInterval="10000" name="">
  <macros/>
  <file>*Security</file>
  <rules>
      <rule name="">
        <match repeatCount="0" repeatInterval="1">.*</match>
        <event>A</event>
        <metrics class="java.util.ArrayList"/>
        <agentAction action=""></agentAction>
      </rule>
  </rules>
</parser>
will there be any events?

Checklist of things:
- agent should have this in it's config:
SubAgent = logwatch
- template with the policy should be applied to node

SubAgent = logwatch

is needed in agent config

It's a bug in the framework that we use, happens with dark Windows theme. We are working on a fix, but this might take some time. Meanwhile you can try light theme or use web ui.

It would be good if you can collect some debug information. If your server is on Linux:

- install netxms-dbg and gdb packages
- stop netxms server:
systemctl stop netxmsd
- launch it under gdb:
gdb --args netxmsd -D 3

then enter run

when it crashes, enter bt and show us the output

----

Meanwhile, you can use another option to monitor windows event logs - by adding log parser policy to a template and applying that template to your nodes. https://www.netxms.org/documentation/adminguide/log-monitoring.html#file-tag
Chances are that this bug will not appear if using this approach (but may be not)

is it this line?:
    if (b->size == 2) h[b] = b[1];
Try running your script via "Execute script" on your PDU node. You can add line like
    println(b);into your script, this way you'll see what exactly is contained in the variables. I'd start with checking of what comes in variable v, it could be that nothing just comes via snmp. Actually, transport->getValue() can return null if there's some problem, so after that line there should be a check for that:

    if (v == null) return null;

What exactly is version of NetXMS ?

It's quite hard to estimate required resources, as they depend not only on the number of nodes, but also on number of interfaces, number of DCIs per node, data collection interval and storage time, number of scripts, number of various events happening and so on. Also, on big systems use of Timescale DB helps, as it minimizes housekeeper resource usage.

Here are some real-life data about tree installations - X, Y and Z. All are using SSD or NVMe storage. On these systems Postgres (non-timescale) was used, running on the same server, so RAM and core numbers cover both NetXMS and DB engine.

Number of datapoints per second was obtained from the DB by this query: SELECT polling_interval, retention_time, count(polling_interval) FROM items GROUP BY polling_interval, retention_time ORDER BY retention_time, polling_interval;
 

	X	Y	Z
Objects	20554	56040	41542
Monitored nodes	2175	2656	5376
Collectible DCIs	34260	185742	372240
Collected datapoints per second	372	979	2290
Stored datapoints, millions	8995	14978	44195
RAM, GB	24	74	160
CPU cores	8	10	20
DB size, GB	237	1500	1500
Bytes per datapoint	53	108	61

Most probably your DCI became unsupported. If you right-click the list of DCIs, there's "show unsupported" in the context menu. Or you can switch to edit mode - there's pen icon above DCI list.

Check that DCI origin is set to script, check the name of the script. If you right-click your node and choose "execute script", you can run your script there to test if it returns some result. Script parameters, e.g. .1.3.6.1.2.1.1.1.0, activePower can be provided via parameters field.

Hi,

The only thing that MIB files provide is textual decoding of OIDs in MIB explorer. MIBs are not related to templates, you have to create that template yourself.

Да, с библиотеками бывает такое

Для мак можно взять 4.5.3, там в плане UI вроде особо не было отличий

Надо смотреть, какую java берет томкал, сейчас требуется 17-я

Можно запустить из-под gdb и когда упадет сделать bt - может быть там что-то будет видно