Zabbix_get showing different value than latest data (Raw data)

[mlrtime]

I have a host with about 747GB of Memory and around 476GB available. I understand how zabbix calculates available memory (https://www.zabbix.com/documentation...ry.size_params). I'm getting a completely different number in raw numbers than zabbix_get

Example:
# zabbix_get -s HOST -p 10050 -k "vm.memory.size[used]"
476883517440

Latest Data:



It looks to me as if the value is too large for the data type in the database

[mlrtime]

Bump, any ideas? Anyone know how I can get the query that this returns, should it be the exact value from zabbix_get?

[Atsushi]

I can't read the text on the attached image file, so I don't know what is different.
Please tell me what OS you are using and what value you want to get. Depending on the OS being used, the size of the available memory can be checked using the following key.

ex.
vm.memory.size[available]

[mlrtime]

Thank you for responding.

Here is the value from zabbix_get
Example:
# zabbix_get -s HOST -p 10050 -k "vm.memory.size[used]"
476883517440

Here is the values I get in HOST->Items->Latest data->Memory (Available memory)->Graph->Values

09/24/2019 07:34:21 AM	6596272128
09/24/2019 07:33:21 AM	6580867072
09/24/2019 07:32:21 AM	6586548224
09/24/2019 07:31:22 AM	6585708544
09/24/2019 07:30:21 AM	6586400768

The values from zabbix_get are not the same as the data shown.

HOST is Centos 7 3.10.0 kernel
Zabbix is Centos 7.6 3.10.0 kernel

I'd like to why there is a discrepancy between zabbix_get values and what I see in the UI

[dimir]

I assume you are sure you are pointing zabbix_get exactly to the same agent that handles host in the graph?

[mlrtime]

Yes, I double checked, I'm obscuring the hostname for simplicity. But yeah, I copy the hostname from the graph page to the zabbix_get and get different values. The graph is 'Memory (Available Memory)' and the get is vm.memory.size[available], I haven't changed the template.

[dimir]

Wait, but available and used are different things, right?

[mlrtime]

Argh, typo in the post (doesn't help my case). Here are the values, wrong value for available in values from UI

# zabbix_get -s HOST -p 10050 -k "vm.memory.size[available]"
789995679744

# zabbix_get -s HOST -p 10050 -k "vm.memory.size[used]"
685048332288


Available memory Values from UI

09/24/2019 08:47:21 AM	6587539456
09/24/2019 08:46:21 AM	6595207168

It doesn't look like used memory is a item captured in the Template OS linux

[dimir]

Just out of curiosity, what if you use zabbix_agentd binary on target host, instead of zabbix_get:

# sudo -u zabbix_agentd -t vm.memory.size[used]
# sudo -u zabbix_agentd -t vm.memory.size[available]

[mlrtime]

Same values as zabbix_get

# zabbix_agentd -t vm.memory.size[available] ; zabbix_agentd -t vm.memory.size[used]
vm.memory.size[available] [u|792433606656]
vm.memory.size[used] [u|682677321728]

[dimir]

Looks like some restrictions issue, is there anything like SELinux running on target host?

[mlrtime]

# getenforce
Disabled

On client and Zabbix Server

The values match on the server and the client, my guess is that the datatype in the db is wrong and truncating. How can I get the exact query that this page uses?

[dimir]

The numbers that UI shows (e. g. 6587539456) would not fit into Integer, so I doubt it's a database issue. Could you run the zabbix_agentd -t commands with sudo -u zabbix as I specifically mentioned? This way we are sure the values are fetched by the same user as the agent is running as.

[DARKINATY]

Yes, I double checked, I'm obscuring the hostname for simplicity. But yeah, I copy the hostname from the graph page to the zabbix_get and get different values. The graph is 'Memory (Available Memory)' and the get is vm.memory.size[available], I haven't changed the template.

I assume you are sure you are pointing zabbix_get exactly to the same agent that handles host in the graph?

[mlrtime]

So this is fixed but I never found the root cause. I upgrade the server and agent to 4.4 and now UI is showing the proper 600GB ram available. I'm assuming this is a zabbix bug