Ad Widget

**jeeva** · 20-01-2022, 10:18

Setup and configure zabbix-agent2 compiled with the Ceph monitoring plugin.

Where does one get this zabbix-agent2 compiled with ceph?

**kuziev** · 28-01-2022, 14:23

Originally posted by setsimmo

I tried to troubleshoot the issues with ceph.osd.stats and found the following:

The default RESTful permissions do not allow the pg dump* commands to be run by the user as created. I couldn't track down where to set permissions in Ceph for RESTful module users, as the users created by "ceph restful create-key" do not show up under other modules.

More details on the RESTful API module can be found here: https://docs.ceph.com/en/latest/mgr/restful/

FIX auth for mgr service and restart mgr service

Code:

ceph auth caps mgr.$id mon 'allow *' osd 'allow *' mds 'allow *'

**bilbolodz** · 09-08-2022, 15:24

I'm trying to figure out how to deal with detection active manager on CEPH and choosing host to ask for API interface. According documentation restful API starts only on manager which is active at that moment. My active manager can be at one of 6 hosts!

**kuziev** · 10-08-2022, 10:38

Originally posted by bilbolodz

I'm trying to figure out how to deal with detection active manager on CEPH and choosing host to ask for API interface. According documentation restful API starts only on manager which is active at that moment. My active manager can be at one of 6 hosts!

nginx config

Code:

upstream cephmgrrestfull {
    server 192.168.99.51:8003  weight=1 max_fails=1 fail_timeout=120 backup;
    server 192.168.99.52:8003  weight=1 max_fails=1 fail_timeout=120 ;
    server 192.168.99.53:8003  weight=2 max_fails=1 fail_timeout=120 backup;
}
server {
    listen 8003 ssl http2 default_server;

    server_name server_domain_or_IP;
    include snippets/self-signed.conf;
    include snippets/ssl-params.conf;

        location / {
            proxy_pass         https://cephmgrrestfull;
        }
}

**bilbolodz** · 10-08-2022, 10:49

Thanks but it require "external to CEPH cluster service" (nginx) running somewhere which I'd like to avoid. It's very strange that CEPH doesn't offer built in service for API redundancy. They already offer HA solution for S3 (HA proxy) and dashboard (HTTP redirect to active manager) so why not for API?

**kuziev** · 10-08-2022, 11:01

Nginx is used as a web server for zabbix, if you have apache you can probably do it on it ( https://httpd.apache.org/docs/2.4/mo..._balancer.html ).

**bilbolodz** · 10-08-2022, 11:07

That's indeed could be a smart idea to fire up HA for CEPH on zabbix server itself!

**bilbolodz** · 18-08-2022, 12:44

Actually my work mate found a better solution which NOT require any additional software:

install zabbix agent 2 on every node which can be running mgr. It's generally a good idea to add it to "usual monitoring" to zabbix
register ALL possible mgr nodes IP under common dns name (e.x. ceph-mgr.intra.blabla.com)
in zabbix create host (e.x ceph-cluster) which represents your ceph cluster and set Interface to "Agent" BUT using DNS name: ceph-mgr.intra.blabla.com
assing Ceph by Zabbix agent 2 template to host ceph-cluster
set {$CEPH.CONNSTRING} to value: https://ceph-mgr.intra.blabla.com:8003 for host ceph-cluster
add ceph-cluster to Hostname directive in zabbix_agent2.conf file on EVERY node can run mgr (Hint: It could be multiple names separated by comas in Hostname directive), restart agent
enjoy working ceph cluster monitoring

**tinomms** · 17-11-2022, 13:01

Hi there folks.

I need some help with this please. I can't find any clear, easy to follow documentation for how to set up this plugin with our CEPH cluster hosted on Proxmox.

My question therefore might have an obvious answer to some, but the documentation doesn't say, so go easy on me. Is the APIkey something that is generated at the command line (if so how?) or is it something that is input (kinda like a password) within the ceph.conf file of the plugin? If none of those then what is it and how is it generated please??

Thanks

Tino

**rmday** · 08-03-2023, 19:46

Hello!

I am a new poster because I am new to zabbix along with ceph. My group upgraded our ceph cluster (vended from Croit), to a new version and since that point, we are noticing that none of the data is getting to our zabbix instance. It seemed to stop the day of the upgrade. As a result, we have OSDs showing as down and others up, but would not see any change in that except direct from the ceph management node.

We have the zabbix agent on all the nodes and the [ceph integration](https://www.zabbix.com/integrations/ceph) had been working for over a year. I am just not sure how to get started troubleshooting this and cannot make a new topic.

Any "get started troubleshooting" help would be appreciated!

**ttyzzx** · 12-09-2023, 12:39

Hi,

It took me somewhat longer than it should have to get this working. I have a couple of suggestions for the README:

The first line under Setup states "Setup and configure zabbix-agent2 compiled with the Ceph monitoring plugin." I was unsure what this meant, I'm using official packages hand have no need to "compile" anything. Perhaps stating the official packages already support this would be less confusing.

It would be useful to reference that this uses the CEPH RESTAPI, and either give a few commands to activate it, or link to the appropriate CEPH page: https://docs.ceph.com/en/latest/mgr/restful/ - Finding this was my light-bulb moment, I was struggling because it was not clear how zabbix and ceph were glued together.

It would be useful to highlight the existence of the configuration file found at /etc/zabbix_agent2.d/plugins.d/ceph.conf.

These things are probably really obvious to those who know about them, but can be real stumbling blocks to those of us coming fresh to this.

Cheers,

Chris

**jartoun** · 08-04-2024, 21:23

Hi everyone,

I am trying to get this template to work with a rook-ceph cluster... I have enabled the dashboard, the restful api and created a api user, however the zabbix template does not work.
I always get {"status": "401 Unauthorized", "detail": "You are not authorized to access that resource", "request_id": "252990b4-55ce-4f6b-8990-06943f624129"}

2 questions...

1. Has anyone actually got this template to work with rook-ceph, in that case, how??
2. How do I change default value strings? for example the template seems to add port number 8003 to the #CONNSTRING macro, is there any way to change that?

BR
jartoun

**bbrendon** · 19-03-2025, 06:29

Since this is using from what I understand the old API and causes memory leaks in ceph-mgr because of ceph bugs, is there a new template for the new API coming? Has anyone started working on it?

HTML Code:

https://tracker.ceph.com/issues/59580
https://www.reddit.com/r/ceph/comments/1ecp6rf/problem_with_restful_module/
https://www.spinics.net/lists/ceph-users/msg77420.html

**lrizzo_inap** · 15-04-2025, 16:03

Hello,

As stated above in multiple messages, a Ceph cluster will have usually multiple managers running at the same time, but zabbix, by relying on the agent to access the ceph rest api, seems to be limited to connecting to only one host at a time; in case of a failure of the node that hosts the manager checked by zabbix, one could/would lose any visibility on the ceph cluster for a prolonged period of time and even relying on DNS like suggested above might require quite some time before any visibility/monitoring is restored.
is there any possibility to load balance the access to the zabbix agent with something like ha proxy?

I am a little inexperienced with haproxy (and with the inner workings of zabbix agent <--> zabbixproxy/zabbix server connectivity) and my first attempt at adding a haproxy service for port 10050 is not working properly even after adding the haproxy IPs in zabbix configs as servers; has anyone attempted something like this successfully or have any suggestion on which settings/configurations might help?

Thanks

**Cond3nz** · 09-07-2025, 13:41

Originally posted by setsimmo

I tried to troubleshoot the issues with ceph.osd.stats and found the following:

The default RESTful permissions do not allow the pg dump* commands to be run by the user as created. I couldn't track down where to set permissions in Ceph for RESTful module users, as the users created by "ceph restful create-key" do not show up under other modules.

More details on the RESTful API module can be found here: https://docs.ceph.com/en/latest/mgr/restful/

Thanks for your information, it just helped me:

Code:

ceph auth caps mgr.{node name} mon "allow profile mgr, allow command 'pg dump' " mds "allow *" osd "allow *"

Ad Widget

Discussion thread for official Zabbix Template Ceph

Comment

Comment

Comment

Comment

Comment

Comment

Comment

Comment

Comment

Comment

Comment

Comment

Comment

Comment

Comment