NetXMS Support Forum

Deadlock is already fixed, so if you don't have performance issues with sqlite, you can keep running server on it.

Best regards,
Victor

Hi!

And what interface is correctly discovered - Vlan-interface104 or Vlan-interface212? Or they both really have same IP address? If not, just delete wrong interface (or  both - system will re-create correct one on next configuration poll).

I didn't see anything wrong with topology polls. I'll add more debug output for map generation in upcoming release - hopefully it will help identify why maps are not created as expected.

Best regards,
Victor

Hi!

How do you install agent? From deb package or from sources?

Best regards,
Victor

Internal state was a bit strange at first glance, this is why I was asking. But SQLite backend explains it. SQLite driver is different from all other drivers and having multiple db connections has no effect on it because internally all queries done from single thread anyway. Because of that systems with sqlite backend had much more chance to encounter deadlock you encountered.

Best regards,
Victor

I probably found the deadlock. I'll fix it in upcoming 1.2.9 release. I have some additional questions though:

1. what database you are using as backend?
2. do you have option EnableMultipleDBConnections set to 1?

Best regards,
Victor

Sorry, I forgot that it was intentional crash

oracle.ddr  в аттаче (собран для версии 1.2.8). Достаточно просто скопировать его в bin каталог агента.

Hi!

First, try to run database check (nxdbmgr check) after server shutdown. If it asks for unbound nodes, answer no to option to bind them and yes for delete them.

Can you give more information about these problematic devices? Ideally would be also server debug log at level 6 when it polls these devices.

Best regards,
Victor

Hi!

Are you using nxadm either manually or from some scripts? According to crash dump server crashed hen processing some command from nxadm.

Best regards,
Victor

Hi!

"system name" is a node property, not a node object name in console. You have couple of options here:

1. if you have correct DNS records, you can turn on DNS node sync by setting server configuration variable SyncNodeNamesWithDNS to 1.

2. If you want to keep node name consistent with system name read from node, you can use the following hook script for configuration polls:


if (($node->snmpSysName != null) && ($node->snmpSysName != ""))
   RenameObject($node, $node->snmpSysName);


Best regards,
Victor

Hi!

Yes, you have to run configure and make again. There should be no risks, it will just build few additional binaries.

Best regards,
Victor

Вообщем баг в инсталлере - забыли положить oracle.ddr У вас 32 или 64 бита система?

Да, я пока первый ответ писал, подумал, что было бы удобно добавить скрипт для фильтрации узлов к автоматическим картам, наподобие фильтра для network discovery. Попробую сделать, по идее должно быть несложно. Тогда можно будет сделать карту L2 узлов из одной IP подсети.

Regarding duplicate IPs: check in console if these interfaces really have same IP addresses as some other interfaces.

Can you please run topology poll from console (Poll -> Topology from context menu on node) on any of switches where error "call to buildL2Topology on object 5008 failed" displayed and post output here?

Best regards,
Victor

It depends on your system settings. Make sure that core dump generation turned on and check where core dumps are stored. There is a lot of instructions for Debian, for example this one: http://mtechease.blogspot.com/2011/10/core-dump-for-debian-mayank-chourey.html

Best regards,
Victor