Monitoring of Kafka cluster and nodes

Last update: November 12 2024

Monitoring of Kafka cluster and nodes

In Kafka's cloud databases, you can track the status of the cluster.

To assess the overall state of the cluster check his status..

For a more detailed analysis, you can:

• see the status of the node cluster — in the form of graphs in the control panel;
• export cluster node metrics in Prometheus format.

You can find out if your disk is about to run out of space by using disk fill notifications.

View cluster status⁠

1. В control panels go to Cloud platform → Databases.
2. In the cluster row, look at the status.

ACTIVE	The cluster is available
CREATING	A cluster is created
UPDATING	Changes are applied to the cluster
RESIZING	The cluster is scalable
ERROR	There's been a mistake, file a ticket
DISK FULL	The disk is full and the cluster is read-only. To make the cluster read and write, disk cleanup or scale the cluster and select a configuration with a larger disk size
DEGRADED	Some nodes in the cluster are unavailable
DELETING	The cluster is being deleted

View the status of the node cluster⁠

1. В control panels go to Cloud platform → Databases.
2. Open the cluster page → tab Monitoring.
3. In the block Monitoring of cluster servers see what's available node cluster metrics.

Cluster node metrics in the control panel⁠

vCPU	How many percent of the node cluster cores are utilized
Load Average	The average value of system load over a period of time. Shows how many processes are processed by the cluster cores. The indicator is presented in the form of three values — for one minute, five minutes and 15 minutes. These values should not be greater than the number of cores on the node
Memory	Memory utilization excluding cache and operating system buffers in percent or gigabytes
Disk	Used disk space in percent or gigabytes. The file system reserves 4% of the disk space for service needs, so the space occupied is greater than the actual space used by this 4%

Export metrics in Prometheus format⁠

Historical information for clusters is not available — metrics are requested only in real time. The list of all metrics that are supported in cloud databases and their description can be seen in the following table Metrics in Prometheus format.

1. Get a token.
2. Get metrics in Prometheus format.

Get a token⁠

The token gives access to the metrics of all clusters project in one bullet.

1. В control panels go to Cloud platform → Databases.

2. Open the cluster page → tab Monitoring.

3. In the block Tokens for Prometheus click Add token. The token will be generated automatically.

4. Copy the token. To do this, in the token row, click .

Get metrics in Prometheus format⁠

Configuration file

1. Add to the Prometheus configuration file:

   scrape_configs:
     - job_name: get-metrics-from-dbaas
       scrape_interval: 1m
       static_configs:
         - targets:
           - '<pool>.dbaas.selcloud.ru'
       scheme: https
       authorization:
         type: Bearer
         credentials: <monitoring_token>

   Specify:

   • <pool> — pool in which the token is valid, e.g. ru-3. The address (URL) depends on the region and pool, you can look in the URL list;
   • <monitoring_token> — a monitoring token that you received Previously.

2. Open a page in your browser where Prometheus-formatted metrics will be available:

   http://<ip_address>:9090/targets

   Specify <ip_address> — IP address where Prometheus is installed.

3. Independently configure monitoring and alerts for database clusters.

CLI

1. Open the CLI.

2. To get the metrics, submit a request:

   curl -L "https://<pool>.dbaas.selcloud.ru/metrics" -H "Authorization: Bearer <monitoring_token>"

   Specify:

   • <pool> — pool in which the token is valid, e.g. ru-3. The address (URL) depends on the region and pool, you can look in the URL list;
   • <monitoring_token> — a monitoring token that you received Previously.

   Available metrics in Prometheus format will appear in the response.

3. Independently configure monitoring and alerts for database clusters.

Metrics in Prometheus format⁠

Metrics in Prometheus format are provided for all clusters. A specific cluster can be found by the database cluster identifier in the label ds_id.

dbaas_memory_percent	Memory utilization excluding cache and operating system buffers (RAM) in percent
dbaas_memory_bytes	Occupied memory excluding cache and operating system buffers (RAM) in bytes
dbaas_oom_count	Number of processes that ended with an error Out of Memory due to lack of RAM
dbaas_cpu	Percent vCPU utilization on database cluster nodes
dbaas_cpu_iowait	I/O waiting time in percent
dbaas_disk_percent	Percentage of disk space occupied. The file system reserves 4% of disk space for service needs, so the occupied space is larger than the actual space used by this 4%
dbaas_disk_bytes	Occupied disk space in bytes. The file system reserves 4% of the disk space for service needs, so the occupied space is larger than the actual space used by this 4%
dbaas_disk_read_iops	Number of read operations per second
dbaas_disk_write_iops	Number of recording operations per second
dbaas_disk_read_bytes	Disk read speed in bytes per second
dbaas_disk_write_bytes	Data write speed to disk in bytes per second
dbaas_node_load1	The average value of system load in one minute. Shows how many processes are processed by the cluster cores
dbaas_node_load5	The average system utilization over five minutes. Shows how many processes are processed by the cluster cores
dbaas_node_load15	The average value of system utilization over 15 minutes. Shows how many processes are processed by the cluster cores
dbaas_network_receive_bytes	Number of bytes received through the network interface
dbaas_network_transmit_bytes	Number of bytes sent through the network interface
dbaas_network_receive_packets	Number of packets received through the network interface per second
dbaas_network_transmit_packets	Number of packets sent through the network interface per second
dbaas_role	Role of the node: 0 — role unknown; 1 — master; 2 — replica

Disk fullness notifications⁠

If the cluster disk is 80% full, a notification will appear in the dashboard and will be sent to the email of the Account Owner and those users subscribed to the the "Services and services" notification category.

If the cluster disk is 95% full or more, the cluster will go into status DISK_FULL and will be read-only. To make the cluster read-write, disk cleanup or scale the cluster and select a configuration with a larger disk size.

Clear the disk⁠

Open a transaction transaction_read_only = no and delete unnecessary data using one of the queries:

• DROP TABLE — deletes the structure (data, privileges, indexes, constraints, triggers). Use when completely deleting a table with data and structure:

  BEGIN;
  SET transaction_read_only = no;
  DROP TABLE table_name;
  COMMIT;

• TRUNCATE TABLE — deletes the contents of the table, but the structure is preserved. Works faster DROP TABLE. Use when deleting all rows of a table while preserving the table structure:

  BEGIN;
  SET transaction_read_only = no;
  TRUNCATE TABLE table_name;
  COMMIT;

• DELETE — use to delete specific strings.

For your information

We do not recommend using the query DELETE FROM table WHERE ... to clean up the disk. This query can create oversized samples on large tables and place them on disk. The remaining free disk space may run out completely, causing problems with Kafka and the need to restore it manually.