Capistrano部署因SSH报错失败：消息过长与文件描述符问题

环境信息 #

• 服务器系统：CentOS
• SSHD版本：OpenSSH_7.4p1, OpenSSL 1.0.2k-fips 26 Jan 2017（无法升级或从源码编译）
• SSHD关键配置：

TCPKeepAlive yes
UseDns no
Protocol 2

问题现象 #

使用Capistrano将项目部署到5台服务器时，多数步骤可正常执行，但在执行restorecon步骤（具体命令：/usr/sbin/restorecon -R /u/apps/my_app/releases/20230724100000）时出现以下错误：

mm_receive_fd: recvmsg: Message too long
mux_master_process_stdio_fwd: failed to receive fd 0 from client
mux_client_request_stdio_fwd: read from master failed: Broken pipe
channel_post_mux_listener: accept: Too many open files
muxclient: master hello exchange failed
mm_receive_fd: recvmsg: Message too long
mux_master_process_stdio_fwd: failed to receive fd 0 from client
mux_client_request_stdio_fwd: read from master failed: Broken pipe

错误根源提示：

Caused by:
Net::SSH::Proxy::ConnectError: command failed: ssh rorra@bastion.mydomain.com -o ConnectTimeout=60 -o ControlMaster=auto -o ControlPath=/tmp/sshrorra@bastion.mydomain.com -W server01.mydomain.com:\22

已尝试的无效操作 #

• 移除ControlMaster配置，问题未解决
• 将ControlMaster设为none，导致Capistrano每一步都要求输入凭据，操作耗时过长且命令繁琐
• 执行命令sudo systemctl restart sshd.service重启SSHD服务，问题依旧存在

内容的提问来源于stack exchange，提问作者rorra