NetXMS Support Forum

Nado smotret' logi servera. Esli bili oshibki, to tam dolzna bit' zapis'. A kak vigljadit komanda, propisannaja v action?

Hi!

Release 1.1.10 planned to be somewhere next week, which will include all these fixes. In the meantime you can just install 1.1.9, replace nxdbmgr.exe, and run database upgrade manually. It will not affect server operation in any way, because this bug is only within database upgrade tool.

Best regards,
Victor

Hi!

Helpdesk state is not an alarm state you see in console. It is special field not visible in console intended for help desk system integration. And alarm state you are interested in is missing from nxalarm I'll add it in next 1.1.x release, and I can create a patched version of nxalarm 1.0.13 if you need.

Best regards,
Victor

Hi!

That's very strange. Just copy your config into file on my test system and it works as expected. What operating system you are using? Are you sure you don't have any non-printable characters in your config file? I also suggest to use fully qualified file name in DBName parameter, otherwise database file will be created in current directory, but it's not a problem here.

Best regards,
Victor

Hi!

It is a bug in nxdbmgr. Attached is patched nxdbmgr.exe for x64 platform. It should work.

Best regards,
Victor

Hi!

It's a percentage, i.e. if the value is 80 it means that 80% of last minute disk was busy. On Windows, it's an average value for performance counter \PhysicalDisk(_Total)\% Disk Time, and on Linux it's calculated from "# of milliseconds spent doing I/Os" field in /proc/diskstats.

Best regards,
Victor

Hi!

Could you please run nxdbmgr upgrade with -t option? I want to see at what query it crashes.

Best regards,
Victor

Process.WkSet represents amount of physical memory taken by the process (it's called RSS - Resident Set Size - on Linux). It is a decision of OS memory management on when remove unused pages from physical memory to swap and if they should be removed at all. For example, active work with dynamic memory allocation and release (and NetXMS do a lot of malloc/free operations) may cause heap to grow and allocate new pages and physical memory, and then after release memory in process heap shrinks and some pages may become unused. If those pages stay untouched for some time, OS memory manager may decide to remove them from physical memory. Or, process memory manager may releases some pages because of heap shrink - then those pages will be removed from process RSS. So, we have two layers of memory management with different implementations - OS memory manager and process heap (malloc, free, and company). They both can affect number of resident pages. You can also monitor Process.VMSize - it gives total amount of memory consumed by process, even if some of that memory is untouched for long time and swapped out. Will be interesting to see if there will be same difference on Windows and on Linux in this parameter.

Best regards,
Victor

As memory usage is not growing, but just going up and down, this is nit caused by memory leak. Most likely this difference caused by different memory management algorithms in Windows and Linux, and/or different database drivers. Also, what parameter you are using - Process.VMSize or Process.WkSet?

Best regards,
Victor

Hi!

Maybe DHCP server is able to update DNS? Otherwise the only external solution I can see is to write a program which will get information from DHCP server somehow and update primary host name/address in NetXMS. Another option may be to add a feature to network discovery in NetXMS to try MAC address matching for newly discovered devices, and if there are already node in database with same MAC address, but different IP address, assume IP address change.

Best regards,
Victor

Hi!

I would suggest calling a script from macro, like this:

1. Create script called "ServiceStateName" in script library:

state = GetEventParameter($event, "currentValue");
switch(state)
{
   case 0:
      return "OK";
   case 1:
      return "error in parameters";
   case 2:
      return "cannot connect to server";
   case 3:
      return "content of does not match";
}
return "UNKNOWN";


2. Use %[ServiceStateName] instead of %<currentValue>

Best regards,
Victor

Hi!

No one complaining so far I agree that this is a flaw in default configuration. I'll change default values of AgentCommandTimeout and portcheck timeout.

Best regards,
Victor

Hi!

Could you please post graphs of memory consumption and results of "show stat" server console command (it will show number of nodes and DCIs)?

Best regards,
Victor

Hi!

Could be either a bug or configuration problem. I suppose that TCP connection timeout (for agent to web server connection) is longer than server timeout for agent communication, and so server just didn't wait long enough to get value for ServiceCheck.HTTP parameter, and because there was no new value, threshold was not triggered. I suggest to 1) specify reasonable but short connection timeout in ServiceCheck.HTTP parameter; 2) ad additional threshold for data collection error.

Best regards,
Victor

Hi!

Graphs looks great! I can imagine a lot of things to enhance (first thing that comes to my mind is to implement scrollable/zoomable graphs, which GraphView supports), but I like the direction. I'm not an Android development guru either, so we are in same position here.

For now, I have committed your changes into subversion, and publish new console on a web site (as an 1.1.9 version). Please send me your desired login name for subversion so you be able to commit changes directly.

Best regards,
Victor