Chapter 7. Managing database replication with Galera

Red Hat OpenStack Platform uses the MariaDB Galera Cluster to manage database replication. Pacemaker runs the Galera service as a bundle set resource that manages the database master/slave status. You can use Galera to test and verify different aspects of the database cluster, such as hostname resolution, cluster integrity, node integrity, and database replication performance.

Similar to other Pacemaker services, you can use the pcs status command to check that the Galera service is running, and on which Controller nodes it is running. For more information about viewing Pacemaker bundle status, see Section 4.3, “Viewing bundle status”.

When you investigate database cluster integrity, each node must meet the following criteria:

  • The node is a part of the correct cluster.
  • The node can write to the cluster.
  • The node can receive queries and write commands from the cluster.
  • The node is connected to other nodes in the cluster.
  • The node is replicating write-sets to tables in the local database.

7.1. Verifying hostname resolution

By default, director binds the Galera resource to a hostname instead of an IP address. Therefore, any problems that prevent hostname resolution, such as misconfigured or failed DNS, might cause Pacemaker to incorrectly manage the Galera resource.

To troubleshoot the MariaDB Galera cluster, you first eliminate any hostname resolution problems, and then you check the write-set replication status on the database of each Controller node. To access the MySQL database, you use the password set by director during the overcloud deployment.

Procedure

  1. From a Controller node, get the MariaDB database root password by running the hiera command.

    $ sudo hiera -c /etc/puppet/hiera.yaml "mysql::server::root_password"
    *[MYSQL-HIERA-PASSWORD]*
  2. Get the name of the MariaDB container that runs on the node.

    $ sudo podman ps | grep -i galera
    a403d96c5026  undercloud.ctlplane.localdomain:8787/rhosp-rhel8/openstack-mariadb:16.0-106            /bin/bash /usr/lo...  3 hours ago  Up 3 hours ago         galera-bundle-podman-0
  3. Get the write-set replication information from the MariaDB database on each node.

    $ sudo podman exec galera-bundle-podman-0 sudo mysql -B --password="[MYSQL-HIERA-PASSWORD]" -e "SHOW GLOBAL STATUS LIKE 'wsrep_%';"
        +----------------------------+----------+
        | Variable_name              | Value    |
        +----------------------------+----------+
        | wsrep_applier_thread_count | 1        |
        | wsrep_apply_oooe           | 0.018672 |
        | wsrep_apply_oool           | 0.000630 |
        | wsrep_apply_window         | 1.021942 |
        | ...                        | ...      |
        +----------------------------+----------+

    Each relevant variable uses the prefix wsrep.

  4. Verify the health and integrity of the MariaDB Galera cluster by first checking that the cluster is reporting the correct number of nodes.

7.2. Checking database cluster integrity

When you investigate problems with the MariaDB Galera Cluster, you can check the integrity of the whole cluster by checking specific wsrep database variables on each Controller node.

Procedure

Run the following command and replace VARIABLE with the wsrep database variable that you want to check:

$ sudo podman exec galera-bundle-podman-0 sudo mysql -B --password="[MYSQL-HIERA-PASSWORD]" -e "SHOW GLOBAL STATUS LIKE 'VARIABLE';"

The following example shows how to view the cluster state UUID of the node:

$ sudo podman exec galera-bundle-podman-0 sudo mysql -B --password="[MYSQL-HIERA-PASSWORD]" -e "SHOW GLOBAL STATUS LIKE 'wsrep_cluster_state_uuid';"

    +--------------------------+--------------------------------------+
    | Variable_name            | Value                                |
    +--------------------------+--------------------------------------+
    | wsrep_cluster_state_uuid | e2c9a15e-5485-11e0-0800-6bbb637e7211 |
    +--------------------------+--------------------------------------+

The following table lists the wsrep database variables that you can use to check cluster integrity.

Table 7.1. Database variables to check for cluster integrity

VariableSummaryDescription

wsrep_cluster_state_uuid

Cluster state UUID

ID of the cluster to which the node belongs. All nodes must have an identical cluster ID. A node with a different ID is not connected to the cluster.

wsrep_cluster_size

Number of nodes in the cluster

You can check this on any node. If the value is less than the actual number of nodes, then some nodes either failed or lost connectivity.

wsrep_cluster_conf_id

Total number of cluster changes

Determines whether the cluster was split to several components, also known as partitions. Partitioning is usually caused by a network failure. All nodes must have an identical value.

In case some nodes report a different wsrep_cluster_conf_id, check the wsrep_cluster_status value to see if the nodes can still write to the cluster (Primary).

wsrep_cluster_status

Primary component status

Determines whether the node can write to the cluster. If the node can write to the cluster, the wsrep_cluster_status value is Primary. Any other value indicates that the node is part of a non-operational partition.

7.3. Checking database node integrity

If you can isolate a Galera cluster problem to a specific node, certain wsrep database variables can indicate the specific problem in the node.

Procedure

Run the following command and replace VARIABLE with the wsrep database variable that you want to check:

$ sudo podman exec galera-bundle-podman-0 sudo mysql -B --password="[MYSQL-HIERA-PASSWORD]" -e "SHOW GLOBAL STATUS LIKE 'VARIABLE';"

The following table lists the wsrep database variables that you can use to check node integrity.

Table 7.2. Database variables to check for node integrity

VariableSummaryDescription

wsrep_ready

Node ability to accept queries

States whether the node can accept write-sets from the cluster. If so, then wsrep_ready is ON.

wsrep_connected

Node network connectivity

States whether the node can connect to other nodes on the network. If so, then wsrep_connected is ON.

wsrep_local_state_comment

Node state

Summarizes the node state. If the node can write to the cluster, then typical values for wsrep_local_state_comment can be Joining, Waiting on SST, Joined, Synced, or Donor.

If the node is part of a non-operational component, then the value of wsrep_local_state_comment is Initialized.

Note
  • The wsrep_connected value can be ON even if the node is connected only to a subset of nodes in the cluster. For example, in case of a cluster partition, the node might be part of a component that cannot write to the cluster. For more information about checking cluster integrity, see Section 7.2, “Checking database cluster integrity”.
  • If the wsrep_connected value is OFF, then the node is not connected to any cluster components.

7.4. Testing database replication performance

If the cluster and the individual nodes are all healthy and stable, you can run performance benchmark tests on the replication throughput by querying specific database variables.

Every time you query one of these variables, a FLUSH STATUS command resets the variable value. To run benchmark tests, you must run multiple queries and analyze the variances. These variances can help you determine how much Flow Control is affecting the cluster’s performance.

Flow Control is a mechanism that the cluster uses to manage replication. When the local received queue exceeds a certain threshold, Flow Control pauses the replication until the queue size goes down. For more information about Flow Control, see Flow Control on the Galera Cluster website.

Procedure

Run the following command and replace VARIABLE with the wsrep database variable that you want to check:

$ sudo podman exec galera-bundle-podman-0 sudo mysql -B --password="[MYSQL-HIERA-PASSWORD]" -e "SHOW STATUS LIKE 'VARIABLE';"

The following table lists the wsrep database variables that you can use to test database replication performance.

Table 7.3. Database variables to check for database replication performance

VariableSummaryUsage

wsrep_local_recv_queue_avg

Average size of the local received write-set queue after the last query.

A value higher than 0.0 indicates that the node cannot apply write-sets as quickly as it receives write-sets, which triggers replication throttling. Check wsrep_local_recv_queue_min and wsrep_local_recv_queue_max for a detailed look at this benchmark.

wsrep_local_send_queue_avg

Average send queue length after the last query.

A value higher than 0.0 indicates a higher likelihood of replication throttling and network throughput problems.

wsrep_local_recv_queue_min and wsrep_local_recv_queue_max

Minimum and maximum size of the local receive queue after the last query.

If the value of wsrep_local_recv_queue_avg is higher than 0.0, you can check these variables to determine the scope of the queue size.

wsrep_flow_control_paused

Fraction of the time that Flow Control paused the node after the last query.

A value higher than 0.0 indicates that Flow Control paused the node. To determine the duration of the pause, multiply the wsrep_flow_control_paused value with the number of seconds between the queries. The optimal value is as close to 0.0 as possible.

For example:

  • If the value of wsrep_flow_control_paused is 0.50 one minute after the last query, then Flow Control paused the node for 30 seconds.
  • If the value of wsrep_flow_control_paused is 1.0 one minute after the last query, then Flow Control paused the node for the entire minute.

wsrep_cert_deps_distance

Average difference between the lowest and highest sequence number (seqno) value that can be applied in parallel

In case of throttling and pausing, this variable indicates how many write-sets on average can be applied in parallel. Compare the value with the wsrep_slave_threads variable to see how many write-sets can actually be applied simultaneously.

wsrep_slave_threads

Number of threads that can be applied simultaneously

You can increase the value of this variable to apply more threads simultaneously, which also increases the value of wsrep_cert_deps_distance. The value of wsrep_slave_threads must not be higher than the number of CPU cores in the node.

For example, if the wsrep_cert_deps_distance value is 20, you can increase the value of wsrep_slave_threads from 2 to 4 to increase the amount of write-sets that the node can apply.

If a problematic node already has an optimal wsrep_slave_threads value, you can exclude the node from the cluster while you investigate possible connectivity issues.