MariaDB high CPU usage

[hbouma]

We are running Fusion 4.1.5 on a Red Hat 7 64bit VM. It is monitoring 6 Nagios XI instances (5.5.1). Between all Nagios XI servers, we are monitoring a total of 17 hosts and 271 services currently. We have 2 cores and 4GB of RAM for the Fusion server, 4 CPU cores and 16GB of RAM for the Nagios XI servers. We have 284 users on each Nagios XI and the Fusion server. The users are mapped.

The mariadb instance is taking up between 35% and 90% of the total CPUs available and everytime that the poll_subsys.php runs, it kills our Nagios XI performance by using 90% of the CPU's.

2018-08-21 10_29_18-Post a new topic • Support Forum • Nagios.png

[cdienger]

Have you gone through https://assets.nagios.com/downloads/nag ... Tuning.pdf at all? I would recommend setting the memory to 1GB and the number of pollers to 6 right of the bat.

[hbouma]

Changes made to Polling memory and number of polls. Other settings were already in place.

CPU on the XI servers still hitting 90% or more when the Fusion poll_subsys.php events occur.

I am seeing the following in the logs from Fusion:

/usr/local/nagiosfusion/var/log/poll_subsys.log

Code: Select all

PHP Fatal error:  Call to a member function debugDumpParams() on a non-object in /usr/local/nagiosfusion/html/includes/classes/db.inc.php on line 589

/usr/local/nagiosfusion/var/log/poll_subsys.7.USERNAME.log

Code: Select all

2018-08-21 16:26:08[s: 0, u: 0] poll_server() unable to poll data for SERVERNAME, u:USERNAME, poll:nagiosxi_bpi
2018-08-21 16:26:08[s: 0, u: 0] poll_server() CHECK YOUR LIVE_DATA_TIMEOUT SETTINGS. IT MAY NEED INCREASED

Polling settings are as follows:

2018-08-21 16_28_57-Administration · Nagios Fusion.png

[cdienger]

Can I assume that SERVERNAME and USERNAME are replacement for actual values or is that is what is actually printed? Does performance improve if you remove that server? Please zip and PM us complete files from today.

[hbouma]

Yes, SERVERNAME and USERNAME are replaced with the actual servername and username in the log file.

There are 2 CPU issues going on.

1: The Mariadb proccess on the Fusion server is always running at up to 90% of the total CPU's.

2: When the poll_subsys.php processes runs, it spikes the CPU on the Nagios XI box to 90% or more. This starts and ends as soon as the Fusion server starts/stops running the poll_subsys.php against that server. Issue occurs on all 6 Nagios XI boxes.

[hbouma]

I am having trouble sending a PM with the log files. There are over 1,000 log files in the 74KB tar.gz file, and the PM page on the forums freezes anytime I attempt to attach the file via IE or Google Chrome.

[hbouma]

Zip files sent via PMs. I had to split it into 3 to get around the timeout issue caused by our security checks scanning the uploaded files.

[hbouma]

Good afternoon.

Any update with this? We are at a standstill implementing Nagios out to our environment until we can clear up this issue.

[cdienger]

Yes, we've received the files and are looking into the matter. I can tell that we'll have to do some digging into this and it may be best to open a ticket instead to make file transfers easier. Please do so at https://support.nagios.com/tickets/

[hbouma]

A ticket has been created.