【最佳实践】解决 Elasticsearch 8.x 滚动升级失败的问题

[2024-10-29T10:23:45,123][WARN ][o.e.t.ClusterConnectionManager] [es-node-02] failed to connect to node [{es-node-01}{...}{8.8.1}]
org.elasticsearch.transport.ConnectTransportException: [es-node-01][10.0.1.10:9300] handshake failed. unexpected remote node [es-node-01]
at org.elasticsearch.transport.TransportService.lambda$connectionValidator$6(TransportService.java:567)
...
Caused by: org.elasticsearch.transport.TransportSerializationException: Failed to deserialize response from handler [ContextRestoreResponseHandler[...]]
at org.elasticsearch.transport.InboundHandler.doHandleResponse(InboundHandler.java:423)
...
Caused by: java.lang.IllegalArgumentException: remote node [{es-node-01}{...}{8.8.1}] is build [a23c735933a8b1c0c3d0873c8ab96349e5101e5e] of version [8.8.1] but this node is build [6db6a780efb93cf7238a877094bd825d9b8b5fe0] of version [8.13.1] which has an incompatible wire format
at org.elasticsearch.transport.TransportService$HandshakeResponse.throwOnIncompatibleBuild(TransportService.java:712)
at org.elasticsearch.transport.TransportService$HandshakeResponse.maybeThrowOnIncompatibleBuild(TransportService.java:697)
at org.elasticsearch.transport.TransportService$HandshakeResponse.<init>(TransportService.java:691)
...

public static class HandshakeResponse extends TransportResponse {
// ...
public HandshakeResponse(StreamInput in) throws IOException {
    super(in);
    version = Version.readVersion(in);
    buildHash = in.readString();
    
    try {
        discoveryNode = new DiscoveryNode(in);
    } catch (Exception e) {
        maybeThrowOnIncompatibleBuild(null, e);
        throw e;
    }
    maybeThrowOnIncompatibleBuild(discoveryNode, null);
    clusterName = new ClusterName(in);
}

private void maybeThrowOnIncompatibleBuild(@Nullable DiscoveryNode node, @Nullable Exception e) {
    if (DiscoveryNode.isServerless() == false && isIncompatibleBuild(version, buildHash)) {
        throwOnIncompatibleBuild(node, e);
    }
}

private static boolean isIncompatibleBuild(Version version, String buildHash) {
    // 关键逻辑：当版本号相同但构建哈希不同时，认为不兼容
    return version == Version.CURRENT && Build.CURRENT.hash().equals(buildHash) == false;
}
}

# 跳过构建哈希严格检查（用于滚动升级）
-Des.serverless_transport=true

systemctl restart elasticsearch

curl -X GET "localhost:9200/_cat/nodes?v"
curl -X GET "localhost:9200/_cluster/health?pretty"

systemctl stop elasticsearch

下载并安装新版本
wget https://artifacts.elastic.co/downloads/elasticsearch/elasticsearch-8.13.1-linux-x86_64.tar.gz
tar -xzf elasticsearch-8.13.1-linux-x86_64.tar.gz
复制配置文件（确保 jvm.options 中包含 -Des.serverless_transport=true）
cp /etc/elasticsearch/elasticsearch.yml /path/to/new/elasticsearch/config/
cp /etc/elasticsearch/jvm.options /path/to/new/elasticsearch/config/

systemctl start elasticsearch

curl -X GET "localhost:9200/_cat/nodes?v"
curl -X GET "localhost:9200/_cat/recovery?v"

watch -n 2 'curl -s "localhost:9200/_cluster/health?pretty"'

# 编辑 jvm.options，注释或删除该行
-Des.serverless_transport=true
# 逐个重启节点
systemctl restart elasticsearch

# 检查所有节点版本
curl -X GET "localhost:9200/_cat/nodes?v&h=name,version,build"
# 输出示例：
name version build
es-node-01 8.13.1 6db6a78
es-node-02 8.13.1 6db6a78
es-node-03 8.13.1 6db6a78
# 检查集群健康状态
curl -X GET "localhost:9200/_cluster/health?pretty"