Zabbix ODBC issue with monitoring Oracle DB

[darkserg89]

Hi Guys,
Zabbix monitors a database oracle 19 via ODBC. There are zabbix server and zabbix proxy, which are doing sql requests to oracle DB via ODBC. Zabbix version is 6.0.21.
Currently there are two issues with monitoring DB via ODBC:
1. Utilization of ODBC pollers grows unstopable during 2-3 days and then zabbix-server/proxy stopes updating odbc items becase of 100% utilization, but if i open any odbc item and click on execute it will gather fresh data from db.
The only solution I found to cronjob to restart server/proxy daemon every day. Here is a graph of poller use:
I've tried to increase amout of odbc pollers, increase timeoutes, updated oracle odbc driver from 12 till 19 version, but there is on result.
2. The second issue, that one of the odbc item stops updating information from oracle, while other are still working. It frozes and starts updating again if I only restart zabbix-proxy. If I go to the settings of frozen item and click 'execute it' it will get fresh info.
Here are examples of two ODBC itmes, they are working with the same oracle db. Both are just simple sql request.

[itsmixa]

Hi!
I have the same problems with ODBC monitoring for Oracle Databases from 10g to 19c. Zabbix server versions are 5.4.12 and 6.2.7.
Utilization of ODBC pollers does not grow, but from time to time various ODBC items stop updating information from Oracle. Meanwile all ODBC poolers are idle.
I do not use Zabbix proxy. And frozen items start updating again if I only restart Zabbix server.
And there are no any traces in Zabbix logfile.

[markfree]

darkserg89, is there no log information about your ODBC items?
Did you check for ODBC queues?

By the way, it seems that your examples were not attached. At least I do not see the graphs you mentioned.

[markosa]

And you are aware of these known issues:
ODBC checks

• MySQL unixODBC driver should not be used with Zabbix server or Zabbix proxy compiled against MariaDB connector library and vice versa, if possible it is also better to avoid using the same connector as the driver due to an upstream bug. Suggested setup:
  
  PostgreSQL, SQLite or Oracle connector → MariaDB or MySQL unixODBC driver MariaDB connector → MariaDB unixODBC driver MySQL connector → MySQL unixODBC driver

• It has been observed that using ODBC checks for monitoring Oracle databases using various versions of Oracle Instant Client for Linux causes Zabbix server to crash. See also: ZBX-18402, ZBX-20803.

[darkserg89]

Hi All, thanks for support
1. Here are ODBC pollers graphs, one from zabbix server, another one from zabbix-server. Basically they both do the same and they are suffering the same, You can see the trend of growing utilization of pollers. It's reseted about 6-6.30 a.m. by restarting the service.
If I won't restart the service items will stop updating information from oracle db.




2. markosa I don't use any odbc except of Oracle.

Currently I use the oracle 19.20 client. Before It's used to 12. The change of the verison hasn't helped so far.
3. markfree I haven't found any odbc errors in logs, but thanks to your advice, I've checked queue and found some issue with DB monitoring.
I will try to increase ODBC one more time, not sure if it helps, because it seams sessions are frozen.

[chiroli]

I have this same problem, when one oracle databases freeze and stop communication, my pollers increase and my connection between zabbix proxy and zabbix server stop. I used to Oracle ODBC 19.14 for monitoring oracle databases but I don´t know How resolve this.

[darkserg89]

Hi chiroli ! I've increased odbc pollers from 10 till 40 and going to monitor the load of pollers. Let you know if it helps, so far not sure.

[darkserg89]

Hi All,
I couldn't resolve the issue, I've tried following options:
1. Change odbc connection options in odbc.ini such as: QueryTimeout, Failover, CloseCursor
2. Increase amout of odbc pollers, but it just increase the time when all odbc will frloze.
I've noticed, that odbc pollers as child process are increasing in process tree.
I've run command: ss -tpn | grep <ip address of oracle db> and it has shown the amount of odbc pollers, which can not finish the task as I susspect:

Is there any idea of what can be done?

[Henry355]

Hi everyone,
I solved this problem with ODBC 19.12 by upgrading the ODBC driver to 23.7 (I also upgraded the instant client at the same time).
RAM has stopped growing for ODBC pools.
Tested on Zabbix in 7.2