mirror of https://github.com/easzlab/kubeasz.git
8.8 KiB
8.8 KiB
Elasticsearch 部署实践
Elasticsearch是目前全文搜索引擎的首选，它可以快速地储存、搜索和分析海量数据；也可以看成是真正分布式的高效数据库集群；Elastic的底层是开源库Lucene；封装并提供了REST API的操作接口。
单节点 docker 测试安装
cat > es-start.sh << EOF
#!/bin/bash
sysctl -w vm.max_map_count=262144
docker run --detach \
--name es01 \
-p 9200:9200 -p 9300:9300 \
-e "discovery.type=single-node" \
-e "bootstrap.memory_lock=true" --ulimit memlock=-1:-1 \
--ulimit nofile=65536:65536 \
--volume /srv/elasticsearch/data:/usr/share/elasticsearch/data \
--volume /srv/elasticsearch/elasticsearch.yml:/usr/share/elasticsearch/config/elasticsearch.yml \
jmgao1983/elasticsearch:6.4.0
EOF
执行sh es-start.sh后，就在本地运行了。
- 验证 docker 镜像运行情况
root@docker-ts:~# docker ps -a
CONTAINER ID IMAGE COMMAND CREATED STATUS PORTS NAMES
171f3fecb596 jmgao1983/elasticsearch:6.4.0 "/usr/local/bin/do..." 2 hours ago Up 2 hours 0.0.0.0:9200->9200/tcp, 0.0.0.0:9300->9300/tcp es01
- 验证 es 健康检查
root@docker-ts:~# curl http://127.0.0.1:9200/_cat/health
epoch timestamp cluster status node.total node.data shards pri relo init unassign pending_tasks max_task_wait_time active_shards_percent
1535523956 06:25:56 docker-es green 1 1 0 0 0 0 0 0 - 100.0%
在 k8s 上部署 Elasticsearch 集群
在生产环境下，Elasticsearch 集群由不同的角色节点组成：
- master 节点：参与主节点选举，不存储数据；建议3个以上，维护整个集群的稳定可靠状态
- data 节点：不参与选主，负责存储数据；主要消耗磁盘，内存
- client 节点：不参与选主，不存储数据；负责处理用户请求，实现请求转发，负载均衡等功能
这里使用helm chart来部署 (https://github.com/helm/charts/tree/master/incubator/elasticsearch)
$ cd /etc/ansible/manifests/es-cluster
# 如果你的helm安装没有启用tls证书，请忽略以下--tls参数
$ helm install --tls --name es-cluster --namespace elastic -f es-values.yaml elasticsearch
- 4.验证 es 集群
# 验证k8s上 es集群状态
$ kubectl get pod,svc -n elastic
NAME READY STATUS RESTARTS AGE
pod/es-cluster-elasticsearch-client-778df74c8f-7fj4k 1/1 Running 0 2m17s
pod/es-cluster-elasticsearch-client-778df74c8f-skh8l 1/1 Running 0 2m3s
pod/es-cluster-elasticsearch-data-0 1/1 Running 0 25m
pod/es-cluster-elasticsearch-data-1 1/1 Running 0 11m
pod/es-cluster-elasticsearch-master-0 1/1 Running 0 25m
pod/es-cluster-elasticsearch-master-1 1/1 Running 0 12m
pod/es-cluster-elasticsearch-master-2 1/1 Running 0 10m
NAME TYPE CLUSTER-IP EXTERNAL-IP PORT(S) AGE
service/es-cluster-elasticsearch-client NodePort 10.68.157.105 <none> 9200:29200/TCP,9300:29300/TCP 25m
service/es-cluster-elasticsearch-discovery ClusterIP None <none> 9300/TCP 25m
# 验证 es集群本身状态
$ curl $NODE_IP:29200/_cat/health
1539335131 09:05:31 es-on-k8s green 7 2 0 0 0 0 0 0 - 100.0%
$ curl $NODE_IP:29200/_cat/indices?v
health status index uuid pri rep docs.count docs.deleted store.size pri.store.size
root@k8s401:/etc/ansible# curl 10.100.97.41:29200/_cat/nodes?
172.31.2.4 27 80 5 0.09 0.11 0.21 mi - es-cluster-elasticsearch-master-0
172.31.1.7 30 97 3 0.39 0.29 0.27 i - es-cluster-elasticsearch-client-778df74c8f-skh8l
172.31.3.7 20 97 3 0.11 0.17 0.18 i - es-cluster-elasticsearch-client-778df74c8f-7fj4k
172.31.1.5 8 97 5 0.39 0.29 0.27 di - es-cluster-elasticsearch-data-0
172.31.2.5 8 80 3 0.09 0.11 0.21 di - es-cluster-elasticsearch-data-1
172.31.1.6 18 97 4 0.39 0.29 0.27 mi - es-cluster-elasticsearch-master-2
172.31.3.6 20 97 4 0.11 0.17 0.18 mi * es-cluster-elasticsearch-master-1
es 性能压测
如上已使用 chart 在 k8s上部署了 7 节点的 elasticsearch 集群；各位应该十分好奇性能怎么样；官方提供了压测工具esrally可以方便的进行性能压测，这里省略安装和测试过程；压测机上执行：
esrally --track=http_logs --target-hosts="$NODE_IP:29200" --pipeline=benchmark-only --report-file=report.md
压测过程需要1-2个小时，部分压测结果如下：
------------------------------------------------------
_______ __ _____
/ ____(_)___ ____ _/ / / ___/_________ ________
/ /_ / / __ \/ __ `/ / \__ \/ ___/ __ \/ ___/ _ \
/ __/ / / / / / /_/ / / ___/ / /__/ /_/ / / / __/
/_/ /_/_/ /_/\__,_/_/ /____/\___/\____/_/ \___/
------------------------------------------------------
| Lap | Metric | Task | Value | Unit |
|------:|-------------------------------------:|-------------:|------------:|--------:|
...
| All | Min Throughput | index-append | 16903.2 | docs/s |
| All | Median Throughput | index-append | 17624.4 | docs/s |
| All | Max Throughput | index-append | 19382.8 | docs/s |
| All | 50th percentile latency | index-append | 1865.74 | ms |
| All | 90th percentile latency | index-append | 3708.04 | ms |
| All | 99th percentile latency | index-append | 6379.49 | ms |
| All | 99.9th percentile latency | index-append | 8389.74 | ms |
| All | 99.99th percentile latency | index-append | 9612.84 | ms |
| All | 100th percentile latency | index-append | 9861.02 | ms |
| All | 50th percentile service time | index-append | 1865.74 | ms |
| All | 90th percentile service time | index-append | 3708.04 | ms |
| All | 99th percentile service time | index-append | 6379.49 | ms |
| All | 99.9th percentile service time | index-append | 8389.74 | ms |
| All | 99.99th percentile service time | index-append | 9612.84 | ms |
| All | 100th percentile service time | index-append | 9861.02 | ms |
| All | error rate | index-append | 0 | % |
| All | Min Throughput | default | 0.66 | ops/s |
| All | Median Throughput | default | 0.66 | ops/s |
| All | Max Throughput | default | 0.66 | ops/s |
| All | 50th percentile latency | default | 770131 | ms |
| All | 90th percentile latency | default | 825511 | ms |
| All | 99th percentile latency | default | 838030 | ms |
| All | 100th percentile latency | default | 839382 | ms |
| All | 50th percentile service time | default | 1539.4 | ms |
| All | 90th percentile service time | default | 1635.39 | ms |
| All | 99th percentile service time | default | 1728.02 | ms |
| All | 100th percentile service time | default | 1736.2 | ms |
| All | error rate | default | 0 | % |
...
从测试结果看：集群的吞吐可以（k8s es-client pod还可以扩展）；延迟略高一些（因为使用了nfs共享存储）；整体效果不错。
中文分词安装
安装 ik 插件即可，可以自定义已安装ik插件的es docker镜像：创建如下 Dockerfile
FROM jmgao1983/elasticsearch:6.4.0
RUN /usr/share/elasticsearch/bin/elasticsearch-plugin install \
--batch https://github.com/medcl/elasticsearch-analysis-ik/releases/download/v6.4.0/elasticsearch-analysis-ik-6.4.0.zip \
&& cp /usr/share/zoneinfo/Asia/Shanghai /etc/localtime