NetXMS Support Forum

У меня работает - можно закрывать хост. В 1.2.13 будет нормально брать русские буквы с SNMP.

Hi!

There was a bug with keepalive messages not being sent. It is fixed in upcoming 1.2.13 release, so hopefully your problem will go away.

Best regards,
Victor

Yes, should not be hard to implement. Added this as feature request: https://www.radensolutions.com/chiliproject/issues/525

Best regards,
Victor

I've added this as feature request: https://www.radensolutions.com/chiliproject/issues/524

Predefined events follows this scheme:

SYS_ - all system events, generated inside server. Default events also have SYS_ prefix (like SYS_THRESHOLD_REACHED).
SNMP_ - events generated from SNMP traps.
DC_ - events generated from threshold violation (DC stands for Data Collection).

In addition, I often use prefix SYSLOG_ for events generated by syslog parser, and some times per-product prefixes as well, like ORACLE_ for all Oracle-related thresholds.

Best regards,
Victor

. is string concatenation operation in NXSL. NXSL does not do macro expansion within strings, so you have to do concatenation. Don't be confused by $ sign - it is valid character in identifiers and has not any special meaning. It's just a naming convention to start names of built-in variables with $. You can name your variable a$ for example.

Best regards,
Victor

Нашел баг похоже - при конвертации строк, полученных через SNMP, выставленный code page игнорируется. Судя по комментариям в коде - это зачем-то было надо. Подумаю как это можно решить. Нет возможности дать мне доступ по SNMP на какое-нибудь устройство, которое русские тексты по SNMP отдает? Это сильно помогло бы в отладке.

Timeout and packet size are optional arguments and can be omitted. There is word "optional" deep in parameter description All following forms are valid:

Icmp.Ping(10.10.10.1)
Icmp.Ping(10.10.10.1,3000)
Icmp.Ping(10.10.10.1,2000,120)
Icmp.Ping(10.10.10.1,,140)

Best regards,
Victor

Сервер не надо со статикой собирать, ничего хорошего из этого не выйдет. Правильней будет так:

./configure --with-server --with-mysql --enable-unicode --disable-xmpp

и как выглядит summary после отработки configure?

Sorry, I misunderstood the situation. Then you can use the following DCI:

Origin = SNMP, Parameter = OID you are using to get IP address

with the following transformation script:


server = FindNodeObject($node, "netxms-server-name");
return AgentReadParameter(server, "Icmp.Ping(" . $1 . ")");


Best regards,
Victor

Quote from: andrey--k on March 05, 2014, 04:34:01 PM
Приведите пожалуйста пример создания Transformatin script для SNMP-table.

Вот пример скрипта для таблицы - списка процессов на Cisco рутере. Колонку TIME из микросекунд переводим в проценты. В аттаче шаблон с етой таблицей, чтобы можно было посмотреть всю конфигурацию целиком.


ciTime = $1->getColumnIndex("TIME");
for(row = 0; row < $1->rowCount; row++)
{
value = $1->get(row, ciTime);
$1->set(row, ciTime, value * 100 / 1000000);
}


Quote from: andrey--k on March 05, 2014, 12:08:26 PM

Quote from: andrey--k on March 04, 2014, 11:09:53 PM
Решил задачу через динамические DCI. Там пишет raw (hex) в БД.

Спасибо.

Решение оказалось неудачным.
Сервер начал потреблять очень много ОЗУ.
Буду пробовать реализовать поддержку CP1251.

Подскажите пожалуйста, deb пакеты в последней версии (1.2.12), для Debian Wheezy, скомпилированы с  --enable-UNICODE ?

Да, с UNICODE.

Hi!

Solution #1 seems correct. You can get value from remote agent with AgentReadParameter function. However, easier would be to use proxy node directly:

On GATEWAY1 a DCI: Origin = Agent, Parameter = Icmp.Ping(ip), Proxy-node = NetXMS-server

No transformation script needed in this case.

Best regards,
Victor

Hi!

In general, all those values should be kept around zero, or on some stable level. It is normal to have occasional spikes on any of them, but average for 10-15 minutes should be low and not growing over time. Long time high or growing values for those DCIs could indicate:

"Configuration poller queue for last minute" - there are not enough configuration pollers or configuration polls take unusually long time. One of the common reason is timeouts because some communications are blocked by firewalls. It is recommended to disable agent and/or SNMP polling for nodes where they are not needed. Since 1.2.13 it is also possible to disable Check Point SNMP agent check globally - this could speed up configuration polling.

"Status poller queue for last minute" - there are not enough status pollers, or status polling interval should be increased.

"Average time to queue DCI for polling for last minute" - usually caused by excessive waits on different internal locks. Sometimes this is a sign of a bug in a server.

"Database writer's request queue for last minute" (all 3 variants) - indicates that there are more outstanding SQL INSERT or UPDATE requests then database can handle. When database by itself can handle more load but database connection is a bottleneck increasing number of database writers will help.

"Data collection poller's request queue for last minute" - server cannot collect data fast enough. This may happen when monitored nodes responds too slowly, or data collection interval set for too low.

Best regards,
Victor

Quote from: andrey--k on March 05, 2014, 11:25:51 AM
Удалил 75000 DCI.
Потребление памяти после запуска уменьшилось с 960 MB до 215 MB.
Нагрузка на ЦПУ упала с 50-60% до 0,5-2%.

What type those DCIs was - SNMP or agent? What was collection interval?

Best regards,
Victor