NetXMS Support Forum

А сервер перезапустили после изменения параметра? Он читается только при старте. Еще я забыл про параметр агента ExecTimeout - время ожидания результата внешней программы. Он тоже по умолчанию 2 секунды. Попробуйте поставить

ExecTimeout = 5000

в nxagentd.conf.

Best regards,
Victor

Quote from: Alex on October 21, 2010, 12:03:40 PM
P.S.> Кстати, можно как-то сделать универсально, чтоб использовать данный вызов из темплейтов, а не прописывать руками на каждый хост? Типа PGW_Act_Calls(%host_ip%)

Можно. Если в темплейте указать макрос %{node_primary_ip}, то при наложении на хост вместо него будет подставлен IP адрес хоста. Все макросы описаны в мануале в разделе 5.3.6.

Best regards,
Victor

Добрый день!

3 раза - это retry. Если сервер получает тайм-аут в ответ на запрос параметра, он сбрасывает соединение, создает новое, и пробует получить параметр еще раз. Всего попыток делается 3. Похоже что скрипт исполняется слишком долго, и сервер не дожидается ответа от агента. Надо либо увеличить тайм-аут для запроса параметров (параметр сервера AgentCommandTimeout, default 2000 миллисекунд), или как-то переделать скрипт, чтобы он укладывался в 2 секунды.

Best regards,
Victor

Hi!

Could you please test agent version 1.0.6 (available at https://www.netxms.org/download/agent_packages/nxagent-1.0.6.exe)? There are some changes in text log file monitoring, and it may work better in your case.

Best regards,
Victor

Pohoze bili ochen' sil'no poporceni timestampi v idata_xx tablicah. Sejchas oni zapolnjajutsja normal'nimi dannimi i situacija ispravljaetsja. Mne na osnove izvestnoj informacii slozno skazat', v chem bila prichina takogo povedenija. Esli sejchas vse ok, to mozno ostavit' kak est'. Chtobi razobratsja, nado izuchat' soderzimoe tablic idata_xx.

It's because server not working with UNICODE texts. All texts coming from client (in UNICODE) converted to server's code page, which is set by CodePage parameter in netxmsd.conf. You can try to set codepage to UTF-8, but it may or may not work.

Best regards,
Victor

Yes, core service have to be restarted.

Best regards,
Victor

Hi!

What database you are using? Also, could you please enable crash dump generation on server and send generated dumps to [email protected]? You can enable crash dump generation by setting two parameters in netxmsd.conf:

CreateCrashDumps = yes
DumpDirectory = some existing directory writable by server process

Best regards,
Victor

Na samom dele eto stranno, poskol'ku v baze server hranit vse timestamp'i v UTC, i perevod v lokal'nuju time zonu proishodit na kliente. T.e. Oracle na etot pereschet vlijat' ne dolzen. A v Last Values vremja vsegda pravil'no otobrazaetsja?

Currently the only option is to recreate nodes. I'll try to add possibility to move node into cluster in 1.0.6 release.

Best regards,
Victor

Dobrij den'!

A konsol' restartili posle smeni nastroek? Poskol'ku pri smene time zoni uze otformatirovannie timestampi ne peredelivajutsja, tol'ko vo vnov' otkrivaemih oknah ili posle refresh'a uchitivaetsja novaja time zona, to mozet vozniknut' takaja situacija. A vot esli eto posle restarta, ili oba okna otkriti posle smeni time zoni, togda eto ochen' strannij bag...

Eto bag v GUI, ispravlju v sledujuschej versii. Mozno pereimenovat' zajdja v properties i vpisav novoe imja v pole "name".

Currently it's a mix. End of sequence flag was introduced later during protocol development, so some lists uses special message, and some uses this flag.

Best regards,
Victor

Hi!

Currently, the only way to lock node's DCI list is to issue CMD_GET_NODE_DCI_LIST, because when I was writing this, I was thinking only about GUI client, which usually need to display current DCI list immediately. Our current intention is to remove all exclusive locks for configuration and replace it with change notifications - this is already done for some parts of server configuration, like use management. Why receiving full DCI list is a problem? You can just ignore it if you don't need it. And for end of list - you don't have to wait a timeout to detect a list end - each NXCP message has a flags field, and in many list responses flag "END OF SEQUENCE" (0x0008) is used to indicate end of list. In case of DCI list, at the end you will get empty message with this flag set.

Best regards,
Victor

Hi!

All %number macros are event-dependent. %5 will be DCI ID for threshold violation events. However, for threshold rearm events DCI ID will be %3 (I know that it's confusing, but unfortunately this cannot be changed without breaking backward compatibility). There should not be any difference on how DCI was created - directly on the node or via template.
Could you please provide more details about your configuration (DCI thresholds configuration, event policy rule)?

Best regards,
Victor