Hi

I am trying to figure out is this galera setup really "production ready" like blob post said. However, I am thinking do you have problem in logic here https://github.com/severalnines/galera-docker-mariadb/blob/master/entrypoint.sh#L167-L203 The idea of that code is that if whole cluster goes down - it will check from all galera nodes which have the latest sequence number. However, when using statefulsets the pods are starting in order - which means that only one pod can report its status -> and it will always get bootstrapped. We should have possibility to start all nodes under statefulset and after that decide which one has the latest seq number, right? Or am I missing something?

When I created docker image with mariadb:latest and applied the mariadb-ss.yaml. I found that the pods are crashing. But works fine when used the default given in the yams with mariadb10.1.

Can you please update or let me know how we can use the latest mariadb-galera stateful deployment.

Also I want to use this deployment for php app and expose the php to external. Any idea will help.

Thanks and regards,
Mark.

The computation of the node to report if there is no higher seqno is wrong as it orders the nodes based on the modifiedIndex only without taking into consideration the highest seqNo.
The proposed fix is to compute it as below:

node_to_bootstrap=$(cat /tmp/out | jq -c '.node.nodes[].nodes[]?' | grep seqno | tr ',:"' ' ' | sort -k5,5r -k11 | head -1 | awk -F'/' '{print $(NF-1)}')

In the report_status script the sequence number is reported under the name wsrep_last_committed, while in the entrypoint the sequence number is checked under the key seqno. This creates the possibility of having two separate clusters if two nodes in a Kubernetes cluster are drained one after the other (galera-2, then galera-1) and uncordoned at the same time.
The suggested fix is to report the sequence number under the key seqno in the report_status script too.

fix_the_reporting_of_seqno_.patch.txt

k8s error message:
Error: failed to start container "galera": Error response from daemon: oci runtime error: container_linux.go:265: starting container process caused "exec: "/entrypoint.sh": permission denied"

The Kubernetes example conveniently glosses over the fact that the etcd-cluster.ymlconfig has zero resilience against failed nodes!
The file deploys 3 etcd pods which act as a cluster, which is great. But if I stop any of these pods on Kubernetes and then try to deploy them again, the cluster doesn't recognize them! They cannot join the cluster anymore, so the "scaling" is just a facade.

In kubernetes, if not utilizing a storage class on AWS, and with StatefulSet with persistent storage, start with creating the PVs and PVCs, it creates 3 PVs and 3PVCs, how can I control which pod of the 3 Pods volume the first pvc?
Use openEBS or nfs storage class is recommend.

Hi

wondering if you would be interested by a PR to add support for etcd+consul as a service discovery backend

thus enabling a hashicorp jobspec example in addition to k8s and docker compose.

?

otherwise i'll start a clean fork

For use etcd

etcd

在 docker 中创建 etcd，需要使用以下环境变量：environment --env -e
ETCD_ENABLE_V2=true

docker compose file:

version: "3.0"

networks:
    net:
        # 网络
        driver: bridge # 桥接模式

volumes:
    data: # 挂载到本地的数据卷名
    data-324:


services:
    etcd-cluster:
        image: bitnami/etcd:3.5 # 镜像
        container_name: etcd-cluster # 容器名 --name
        restart: always # 总是重启
        networks:
            - net # 使用的网络 --network
        ports:
            # 端口映射 -p
            - ${your_private_ip}:2379:2379 # e.g.  - 2379:2379
            - ${your_private_ip}:2380:2380 # e.g.  - 2380:2380
        environment:
            # 环境变量 --env
            - ALLOW_NONE_AUTHENTICATION=yes # 允许不用密码登录
            - ETCD_ENABLE_V2=true # 这个参数是成功的关键 / This parameter is the key to success.
            - ETCD_NAME=etcd-node1 # etcd 的名字
            - ETCD_INITIAL_ADVERTISE_PEER_URLS=http://${node1_domain_or_ip}:2380 # 列出这个成员的伙伴 URL 以便通告给集群的其他成员
            - ETCD_LISTEN_PEER_URLS=http://${your_private_ip}:2380 # 用于监听伙伴通讯的URL列表 e.g. http://0.0.0.0:2380
            - ETCD_LISTEN_CLIENT_URLS=http://${your_private_ip}:2379 # 用于监听客户端通讯的URL列表 e.g. http://0.0.0.0:2379
            - ETCD_ADVERTISE_CLIENT_URLS=http://${node1_domain_or_ip}:2379 # 列出这个成员的客户端URL，通告给集群中的其他成员
            - ETCD_INITIAL_CLUSTER_TOKEN=etcd-cluster # 在启动期间用于 etcd 集群的初始化集群记号
            - ETCD_INITIAL_CLUSTER=etcd-node1=http://${node1_domain_or_ip}:2380,etcd-node2=http://${node2_domain_or_ip}:2380 # 为启动初始化集群配置
            - ETCD_INITIAL_CLUSTER_STATE=existing # 初始化集群状态
        volumes:
            - data:/bitnami/etcd # 挂载的数据卷

mariadb-galera

在多网卡环境中需要指定 hostname extra_hosts multi-interface
docker compose file:

version: "3.8"

# networks:
#     net:


volumes:
    data:


services:

    mariadb-galera:
        image: severalnines/mariadb
        container_name: mariadb-galera
        extra_hosts:
            - node1.example.org:${galera_node1_IP}
        hostname: node1.example.org
        environment:
            - CLUSTER_NAME=mariadb-galera
            - MYSQL_ROOT_PASSWORD=your_PASS
            - XTRABACKUP_PASSWORD=mypassword
            - MYSQL_USER=drone
            - MYSQL_USER_PASSWORD=your_PASS_2
            - MYSQL_USER_DATABASE=drone
            - DISCOVERY_SERVICE=${etcd_node1_IP}:2379,${etcd_node2_IP}:2379
        volumes:
            - data:/var/lib/mysql
        network_mode: host
        # networks:
        #     net:
        # ports:
        #     - ${galera_node1_IP}:3306:3306
        #     - ${galera_node1_IP}:4567:4567
        #     - ${galera_node1_IP}.59:4568:4568
        restart: unless-stopped

Mariadb 10.1 is quite old now and for example Flyway refuses to work with 10.1 now because it's to old, forcing you to use the commercial version with less version constraints.
I do an update to 10.2 in my chain, but going higher would require forking this repo, which everyone would need to do.
Could you update to a newer mariadb version, please? :-)

show status like 'wsrep_cluster_size';
+--------------------+-------+
| Variable_name | Value |
+--------------------+-------+
| wsrep_cluster_size | 1 |
+--------------------+-------+

Databases does not sync.

As a student, I do not have multiple servers, and I always fail to build multiple database clusters based on a single server. I hope that a single "docker-comemess. yml" file can teach me, so that I can learn the configuration principle from a file, thank you.

I think there is a issue computing the running nodes at lines: https://github.com/severalnines/galera-docker-mariadb/blob/master/entrypoint.sh#L250#L256 . The test should be done on the running_nodes3 variable, not running_nodes2.

What is the purpose of the preflight_check function？

Hi,

I don't know if it's because of the docker image or what, but I can't add my scripts to initialize the database.

Hi,
I am using 3 VMS running on virtual box each having 2gb of RAM allocated. I have installed docker v 18.03 on it and created a swarm.
Now im trying to deploy the predictable instance of Mariadb using the global.yml, but the instance isnt coming up, it keeps on failing and re-starting, in the service logs i keep seeing MySQL init process failed.

ServiceLog.txt