Set up the WEKAmon external monitoring
The WEKAmon is an external monitoring package that provides enriched monitoring capabilities using the Grafana and Prometheus tools.
Last updated
The WEKAmon is an external monitoring package that provides enriched monitoring capabilities using the Grafana and Prometheus tools.
Last updated
WEKA provides an external monitoring package named WEKAmon. The package implements the well-known Grafana dashboard with Prometheus, which provides a central monitoring dashboard of metrics, logs, alerts, and statistics with enriched capabilities.
The package also includes the following components:
Exporter: The Exporter gets the data from the Weka cluster and sends the data to Prometheus.
Quota Export: The Quota Export manages the quotas and sends the data to Prometheus.
Alert Manager: The Alert Manger alerts users through an SMTP server when they reach their soft quota limits.
You can set up the Weka-mon package regardless of the data monitoring provided by the Weka GUI.
One of the advantages of setting up the WEKAmon package is that if you already use the Grafana and Prometheus tools for monitoring other products, you can integrate these tools with WEKA to correlate and display monitoring information from all your products on the same dashboard.
Setting up a dedicated management server (or VM) for the installation is recommended.
4 cores
16 GB RAM
50 GB / partition (for the root)
50 GB /opt/ partition (for WEKAmon installation)
1 Gbps network
Docker is the recommended container for the WEKAmon setup. To use Docker, the following must be installed on the dedicated management server:
docker-ce
docker-compose
or docker-compose-plugin
depending on the existing operating system.
For the Docker installation instructions, see the Docker website.
To enable communication between the management server and the Weka cluster, the security token is required in the auth-token.json file.
Create the directory ~/.weka
in the WEKAmon installation directory.
Generate the auth-token.json
file and save it in the ~/.weka
directory you created. See the Obtain authentication tokens topic.
It is highly recommended to create a local user with ReadOnly privilege just for the WEKAmon package and use it for cluster communications. To create a local user, see the Create local users topic.
Obtain the WEKAmon package: Obtain the WEKAmon package from the GitHub repository by downloading or cloning.
Run the install.sh script: The script creates a few directories and sets their permissions.
Edit the export.yml file: The export.yml
file contains the WEKAmon and the exporter configuration. Customize the file according to your actual Weka deployment.
Edit the quota-export.yml file: The quota-export.yml
file contains the configuration of the quota-export container. Customize the file according to your actual Weka deployment.
Start the docker-compose containers: Once done, you can connect to Grafana on port 3000 of the management server running the docker containers.
The Weka-mon package resides on the GitHub repository. Obtain the WEKAmon package using one of the following methods:
On the latest release section, select the Source Code link to download.
Copy the downloaded source code to the dedicated management server or VM and unpack it.
Run the following commands to clone the WEKAmon package from GitHub:
The install.sh
script creates a few directories and sets their permissions.
Run the following command:
The WEKAmon and the exporter configuration are defined in the export.yml
file.
Change the directory to weka-mon
and open the export.yml
file.
In the cluster section under the hosts list, replace the hostnames with the actual hostnames/IP addresses of the Weka containers (up to three would be sufficient). Ensure the hostnames are mapped to the IP addresses in /etc/hosts.
3. Optional. In the exporter section, customize the values according to your preferences. For details, see the Exporter configuration options topic below.
4. Optional. Add custom panels to Grafana containing other metrics.
All other settings in the export.yml
file have pre-defined defaults that do not need modification to work with WEKAmon. All the configurable items are defined but marked as comments by an asterisk (#).
To add custom panels to Grafana containing other metrics from the cluster, you can remove the asterisk from the required metrics (uncomment).
Example: In the following snippet of the export.yml
, to enable getting the FILEATOMICOPEN_OPS statistic, remove the #
character at the beginning of the line.
If the statistic you want to get is in a Category that is commented out, also uncomment the Category line (the first line in the example). Conversely, insert the # character at the beginning of the line to stop getting a statistic.
The WEKAmon deployment includes a dedicated container named quota-export. The container includes an Alert Manager that emails users when they reach their soft quota.
The configuration of the quota-export container is defined in the quota-export.yml
file.
Go to the weka-mon
directory and open the quota-export.yml
file.
Specify the same hosts as you specified in the export.yml file
(see above).
The configuration of the Alert Manager is defined in the alertmanager.yml
file found in the etc_alertmanager
directory. It contains details about the SMTP server, user email addresses, quotas, and alert rules. To set this file, contact the Customer Success Team.
Run the following command:
2. Verify that the containers are running using the following command:
Example:
If the status of the containers is not up, check the logs and troubleshoot accordingly. To check the logs, run the following command:
Once all containers run, you can connect to Grafana on port 3000 of the management server running the docker containers. The default credentials for Grafana are admin/admin
.
If you already have Grafana and Prometheus running in your environment, you only need to run the exporter and add it to the Prometheus configuration.
Follow the steps in the 1. Obtain the WEKAmon package section.
In the Grafana application, import the dashboard JSON
files from the directory weka-mon/var_lib_grafana/dashboards
. For instructions, see the Import dashboard topic in Grafana documentation.
Follow steps 3 and 4 in the above sections.
Do one of the following:
Run the exporter in the docker container (if you have a docker, this is the simple method).
Run the exporter as a compiled binary (if you do not have a docker, use this option)
Run the exporter as a Python script (requires installing a few Python Modules from PyPi).
Get and run the container (the export.yml
configuration file is already edited).
The following example maps the export.yml
configuration file in several volumes in the container:
~/.weka directory
to enable the container to read the authentication file.
/dev/log
to enable entries in the Syslog.
/etc/hosts
to enable the hostname resolution (a DNS can also be used, if exists in the docker environment).
There are more options; you can run the command with-help
or -h
for a full description.
Go to https://github.com/weka/export/releases and download the tarball from the latest release.
Copy this file to the management server or VM.
Run the exporter as follows (for the description of the command-line parameters, see the Exporter section parameters):
Do one of the following:
Run git clone https://github.com/weka/export
Go to https://github.com/weka/export/releases and download the source tarball.
Install the required python modules by running the following command:
pip3 install -r requirements.txt
Run the exporter (for the description of the command-line parameters, see the Exporter section parameters):
The exporter section defines the program behavior.
listen_port
The Prometheus listening port. Do not modify this port unless you modify the Prometheus configuration.
loki_host
If using the Weka-mon setup, do not modify the hostname. Leave blank to disable sending events to Loki.
loki_port
If using the Weka-mon setup, do not modify the port.
timeout
The max time in seconds to wait for an API call to return. The default value is sufficient for most purposes.
max_procs
and max_threads_per_proc
Define the scaling behavior. If the number of hosts (servers and clients) exceeds max_threads_per_proc
, the exporter runs more processes accordingly.
Example: a cluster with 80 Weka servers and 200 compute nodes (aka clients) has 280 hosts. With the default max_threads_per_proc
of 100, it runs 3 processes (280 / 100 ~ 3).
It's recommended to have 1 available core per process. In this cluster example, deploy at least 4 available cores on the server/VM.
backends_only
Run only on the Weka backend hosts
The exporter always tries to allocate one host per thread but does not exceed the maximum processes specified in the max_procs
parameter. In a cluster with 1000 hosts, it doubles or triples up the hosts on the threads.
Example: In a cluster with 3000 hosts, max_procs
= 8, and max_threads_per_proc
= 100, only 8 processes running. Each process with 100 threads, but there are close to 4 hosts serviced per thread instead of the default 1 host.