NetXMS Support Forum

MariaDB [(none)]> show status where variable_name = "innodb_page_size";
+------------------+-------+
| Variable_name    | Value |
+------------------+-------+
| Innodb_page_size | 16384 |
+------------------+-------+


The database was old. I upgraded to debian 10.

I just did a dump, drop, create and restored of the database and known nxdbmgr upgrade is working and netxmsd is running again.

My MariaDB standard debian. No parameters are changed.

In the standaard debian conf innodb_page_size is not set.

- According to de mariaDB documentation innodb_page_size can only be changed for a new databases (https://mariadb.com/kb/en/how-to-change-innodb_page_size/)

- increasing innodb_page_size is bad for tables with small record size and a lot of inserts like the tables that store the values.


PS: I think it is wise to remove 36.4 from the distributions. I just did a apt-get update && apt-get upgrade and now the netxms server is broken.

While upgrading the database to 36.4 i get the following error:

Upgrading from version 36.3 to 36.4
SQL query failed (Row size too large. The maximum row size for the used table type, not counting BLOBs, is 8126. This includes storage overhead, check the manual. You have to change some columns to TEXT or BLOBs):
ALTER TABLE nodes ADD agent_cert_mapping_method char(1)
Rolling back last stage due to upgrade errors...
Database upgrade failed

I am running Debian 10 and MariaDB (10.3.27)

Version 1.2.16.

The strange thing is that sometimes it runs days without any messages and sometime +100 a day.

It is very simple

MasterServers = XXX.YYY.ZZZ
ConfigIncludeDir = C:\NetXMS\etc\nxagentd.conf.d
LogFile = C:\NetXMS\log\debug.log
FileStore = C:\NetXMS\var
SubAgent = logwatch.nsm
SubAgent = winperf.nsm
DebugLevel = 9

I have 100+ warning 'Communication session broken: A request to send or receive data was disallowed because the socket had already been shut down in that direction with a previous shutdown call.' in the system event log. I have this problem on 2 different systems. One windows 2008 and the other 2012.
On the 2012 i changed to configuration to create a debug log.

This is a section from the debug log:

[18-Oct-2014 16:59:39.703] [DEBUG] [session:0] Received message CMD_GET_PARAMETER
[18-Oct-2014 16:59:39.703] [DEBUG] [session:0] Requesting parameter "Net.Interface.Link(21)"
[18-Oct-2014 16:59:39.703] [DEBUG] [session:0] GetParameterValue(): result is 0 (SUCCESS)
[18-Oct-2014 16:59:39.703] [DEBUG] [session:0] Sending message CMD_REQUEST_COMPLETED (size 48)
[18-Oct-2014 16:59:39.703] [DEBUG] [session:0] Message dump:
  ** 0041000000000038000083C100000001
  ** 00000055018F87AB0000001A00530079
  ** 007300740065006D002E005500700074
  ** 0069006D00650002
  ** code=0x0041 (CMD_GET_PARAMETER) flags=0x0000 id=33729 size=56 numFields=1
  ** [    85] STRING "System.Uptime"

[18-Oct-2014 16:59:39.703] [DEBUG] [session:0] Received message CMD_GET_PARAMETER
[18-Oct-2014 16:59:39.703] [DEBUG] [session:0] Requesting parameter "System.Uptime"
[18-Oct-2014 16:59:39.703] [DEBUG] [session:0] GetParameterValue(): result is 0 (SUCCESS)
[18-Oct-2014 16:59:39.703] [DEBUG] [session:0] Sending message CMD_REQUEST_COMPLETED (size 56)
[18-Oct-2014 16:59:39.750] [DEBUG] [session:0] Message dump:
  ** 0041000000000038000083C200000001
  ** 00000055017200660000001800410067
  ** 0065006E0074002E0055007000740069
  ** 006D006502010002
  ** code=0x0041 (CMD_GET_PARAMETER) flags=0x0000 id=33730 size=56 numFields=1
  ** [    85] STRING "Agent.Uptime"

[18-Oct-2014 16:59:39.750] [DEBUG] [session:0] Received message CMD_GET_PARAMETER
[18-Oct-2014 16:59:39.750] [DEBUG] [session:0] Requesting parameter "Agent.Uptime"
[18-Oct-2014 16:59:39.750] [DEBUG] [session:0] GetParameterValue(): result is 0 (SUCCESS)
[18-Oct-2014 16:59:39.750] [DEBUG] [session:0] Sending message CMD_REQUEST_COMPLETED (size 56)
[18-Oct-2014 17:00:40.199] [DEBUG] [session:0] Session disconnected by watchdog (last activity timestamp is 1413644379)
[18-Oct-2014 17:00:40.199] [DEBUG] [session:0] CommSession::disconnect()
[18-Oct-2014 17:00:40.199] [WARN ] Communication session broken: A request to send or receive data was disallowed because the socket had already been shut down in that direction with a previous shutdown call.
[18-Oct-2014 17:00:40.199] [DEBUG] [session:0] Session with 192.168.202.11 closed
[18-Oct-2014 17:00:44.777] [DEBUG] Incoming connection from 192.168.202.11
[18-Oct-2014 17:00:44.777] [DEBUG] Connection from 192.168.202.11 accepted
[18-Oct-2014 17:00:44.792] [DEBUG] [session:0] Message dump:
  ** 00B60020000000100000000000000000
  ** code=0x00B6 (CMD_GET_NXCP_CAPS) flags=0x0020 id=0 size=16 numFields=0

[18-Oct-2014 17:00:44.792] [DEBUG] [session:0] Received control message CMD_GET_NXCP_CAPS
[18-Oct-2014 17:00:44.792] [DEBUG] [session:0] Sending message CMD_NXCP_CAPS (size 16)
[18-Oct-2014 17:00:44.792] [DEBUG] [session:0] Message dump:
  ** 0003000000000010000083C400000000
  ** code=0x0003 (CMD_KEEPALIVE) flags=0x0000 id=33732 size=16 numFields=0


It seems that the connect was closed by the watchdog (red line) but way?

When I unpack the packages (dpkg –x) and use 'ldd –v' to see which library is used it seems to reference libcrypto.so.1.0.0 (part of libssl1.0.0).

Maybe I am wrong but it seems the reference in the .deb file to libssl0.9.8 is wrong.

I can't install netxms-base_1.2.12_amd64.deb (http://www.netxms.com/apt/dists/wheezy/main/binary-amd64/netxms-base_1.2.12_amd64.deb) on debian wheezy.

I get the following error:
netxms-base depends on libssl0.9.8; however:
  Package libssl0.9.8 is not installed.

I have the idea that the .deb package mistakenly reference libssl0.9.8.
After unpacking the package the libraries in the pack reference libcrypto.so.1.0.0

First a look at how Linux is doing it. The output of 'ip address show' has the following structure:

+ interface 0
        +--- IP address A
        +--- IP address B
+ interface 1
        +--- IP address C
+ interface 2
        +--- IP address D

Option 2 is the same as the current situation. You start with the IP address so what to do if a interface has no IP adres? Second a IP address is normaly only link to 1 interface. If multiple interfaces are bound to 1 it wil also show as 1 bounded interface.

Why do you have only network objects under a node and not somthing like this?

NODE
+ CPU
+ MEM
+ DISK
    + DISK0
    + DISK1
    + DISK2
+ Interface
    + eth_inet0
    + eth_vlan
        + eth_dmz
        + eth_lan
        + eth_inet1
+ FileSystem
    + root
    + var
+ Network addresses
    + IP Address 1
    + IP Address 2
+ Services
    + File server
    + Web server
    + Mail server

PS: I am a newbe to NetXMS i just installed it 1 month a go. But it really looks and feels great. Also the Android APP!

Best regards,
Peter

Quote from: Marco Incalcaterra on March 18, 2013, 04:36:26 PM

Quote from: peterjn on March 18, 2013, 11:54:27 AMA NIC can be down or broken and the system will still response to the IP address linked that NIC (using a other NIC).

Hera I'm a bit in trouble to follow you, maybe depends on definition of "down" and "broken"

Example a linux box has 2 NIC, NIC 1 has IP address A and NIC 2 is has IP address B. I have PC connect to NIC 2 and NIC 1 is not connected (down). When I ping IP address A from the PC I wil get a reply. So a ping to IP address A says nothing about the status of NIC 1. It only tells that IP address A is configured on the box.
Windows is a little more complex. When a NIC driver report that the link is down the IP address is removed. But when a NIC is broken (reports false UP) windows wil do the same as linux.

Quote from: Marco Incalcaterra on March 18, 2013, 04:36:26 PM

Quote from: peterjn on March 18, 2013, 11:54:27 AM
One of the reason for my original posting was that we have problems with a NIC linked to the software Hyper-V switch and I want to monitor this. Currently this is not possible.

Best regards,
Peter

In my Hyper-V I have agents running in the VMs, if there is a communication problem with the NIC of the server I'm unable to communicate with the VMs. I use that as first indicator, may be that could be useful for a basic diagnosis even in your case.

Best regards,
Marco

I have the idee that the before the NIC is going down the speed is reduced and i wanted to test this hypothesis. The link speed in the VMs is always 10Gbps

The current rule in NetXMS that there is a logical NIC based on one IP adres is to simple. NIC can be linked to multple IP addresses, Multiple NIC can be bound to 1 NIC, NIC can have multple virtual VLAN NIC etc.

My request to Victor Kirhenshtein is treat a NIC as a device. a IP address is something of the OS and there is no hard relationship between IP address and NIC. Under windows there are a lot of (unused) virtual NICs so hidding a NIC from de UI is desirable.

Best regards,
Peter

Quote from: Marco Incalcaterra on March 12, 2013, 11:11:07 AM
In my opinion, since the MAC address is unique, each interface should be identified by the MAC address and all the IPs associated to that interface should be reported as a list, possibily with corresponding net mask and gateway (if present).

Best regards,
Marco

MAC address is a layer 2 address and is only valid for layer 2 endpoints. So NIC used in a layer 2 switches don't need to have a MAC addresses.
A default gateway is just a route to 0.0.0.0/0 and is like any other routes and is system based.
In windows a route is (simple) destination based routing. Linux has iproute2 and routing can be complex with routing rules.

Quote from: Victor Kirhenshtein on March 11, 2013, 08:44:29 PM
NetXMS counts each logical interface as separate object. I'm thinking about changing this to have multiple IP addresses on one interface, but decision is not made yet.

Best regards,
Victor

In every operating system. A NIC is a device that can be linked to one or more protocol stacks, software layer 2 switch, etc.

A IP address is given to the IP stack and not to a NIC! The IP stack needs to known how to route outing packets and answer to incoming packets there for the IP stacks links the IP address to a NIC. Pinging a IP address says nothing of the NIC linked to that address is up and running. A NIC can be down or broken and the system will still response to the IP address linked that NIC (using a other NIC). This is because the IP addresses is given to the IP stack and not the NIC.

With Hyper-V at least 1 NIC is assigned to a software layer 2 switch. Virtual VM-NICs in are linked to the software switch. Also a host virtual NIC can be created and linked to the software switch. So the real NIC has no IP address and its MAC address is not used. But it is a real NIC with a status, link speed, etc.

One of the reason for my original posting was that we have problems with a NIC linked to the software Hyper-V switch and I want to monitor this. Currently this is not possible.

Best regards,
Peter

I have 2 problems

Windows 2008 R2 Hyper-V host (nsagent 1.2.6 installed)
If a nic is used by the Hyper-V network switch (only bound to Microsoft Virtual Network Switch Protocol). It is not showing up in the console.

Linux firewall (snmpd)
If a nic has multple IP addresses I get the same nic multiple times.

Is there a fix for this or workaround?