dash 4.0.7 → 4.0.8

checksums.yaml

@@ -1,7 +1,7 @@
 ---
 SHA256:
-  metadata.gz: d13e0fdb207715506d85fa85be2a719ca70640130f22ca948c075fd787b3dd88
-  data.tar.gz: cc42a4346178f4ffbcf7b7b2f03820e49f99e9d54c78ab864493a0c0b8b214e4
+  metadata.gz: a417156134dd79a4724f87192f70baa55fdbc4478a115e021845e3742bacee88
+  data.tar.gz: ebdbb8d7f192b692c31c8cdaa9834584c01dec3653b744b1a481672fee179d28
 SHA512:
-  metadata.gz: cdf2135081cc1acc7de1d73d7c7156f93262ad82a5848a6db6aaf484bf4859ed3e529fede60ac0c67df99ec021acdd2cbf6a9175982af8f31873afa03cc2113b
-  data.tar.gz: 30eae7e9f41f2b216be4dd8d16ce583f5a8e2392d3b05b3ad15fcd4ba40645441c05c39a643cca2bf01876ee28aec2fa668a7ffe2a474e27c69b46edc341dbc6
+  metadata.gz: 3b263730b7ccf4eeeb857304477945fd9d2476bdf52cdb30a964b12b09143b36b5f585ae672d776eac29a6e009606add151aa6e9dd98385f9146980bbc40a9db
+  data.tar.gz: a3ebb3e0141a0f692bd70205aa3d438f9a3cbd1022d1b92c229bf1daf19294c4826a7420c0fefbb077f5a6d6cb46016ee6ab3926bb08f11383bb7a6eb0aaa23e

data/lib/dash/configuration/docs/ssh.yml

@@ -79,3 +79,11 @@ ssh:
   # host or tunnel that does not support agent forwarding (for example,
   # Cloudflare Access for Infrastructure with SSH).
   forward_agent: false
+
+  # Connect timeout
+  #
+  # Seconds to wait for the TCP connection and the SSH handshake with a
+  # host before giving up with `Net::SSH::ConnectionTimeout`. Defaults to
+  # 30. Without a bound, a host that accepts the connection but never
+  # answers the handshake hangs the deploy indefinitely.
+  connect_timeout: 30

data/lib/dash/configuration/docs/sshkit.yml

@@ -23,7 +23,8 @@ sshkit:
   #
   # A pooled connection that the network dropped while idle is evicted and the command
   # retried once on a fresh connection, so a long wait elsewhere (a server-lock poll, a
-  # loadbalancer reboot) does not fail the next command on another host.
+  # loadbalancer reboot) does not fail the next command on another host. A host that
+  # stops answering keepalives mid-command is not retried; that deploy fails.
   pool_idle_timeout: 300
 
   # DNS retry settings

data/lib/dash/configuration/ssh.rb

@@ -57,8 +57,15 @@ class Dash::Configuration::Ssh
     ssh_config["forward_agent"]
   end
 
+  # net-ssh only bounds the TCP connect and the version/key exchange when
+  # `timeout` is set. Without it, a host that accepts the TCP connection but
+  # never finishes the handshake blocks the deploy forever.
+  def connect_timeout
+    ssh_config.fetch("connect_timeout", 30)
+  end
+
   def options
-    { user: user, port: port, proxy: proxy, logger: logger, keepalive: true, keepalive_interval: 30, keys_only: keys_only, keys: keys, key_data: key_data, config: config, forward_agent: forward_agent }.compact
+    { user: user, port: port, proxy: proxy, logger: logger, keepalive: true, keepalive_interval: 30, keys_only: keys_only, keys: keys, key_data: key_data, config: config, forward_agent: forward_agent, timeout: connect_timeout }.compact
   end
 
   def to_h

data/lib/dash/sshkit_with_ext.rb

@@ -181,8 +181,15 @@ class SSHKit::Backend::Netssh
   # The block is rerun, so a drop *mid-command* runs the command twice. That
   # is accepted: the errors here are the idle-drop ones, where nothing reached
   # the server, and the alternative is the deploy failing on the next host.
+  #
+  # Net::SSH::Timeout (a Disconnect subclass) is deliberately excluded. net-ssh
+  # raises it when a host has ignored `keepalive_maxcount` keepalives in a row,
+  # which is a host that stopped answering mid-command, not a dropped idle
+  # socket. Retrying that reconnects to a host that has just proven unresponsive
+  # and, if it accepts the connection but never finishes the handshake or the
+  # command, hangs the deploy with no further guard. Let it fail.
   module ReconnectOnStaleConnection
-    STALE_CONNECTION_ERRORS = [ Errno::ECONNRESET, Errno::EPIPE, Net::SSH::Disconnect, Net::SSH::Timeout ].freeze
+    STALE_CONNECTION_ERRORS = [ Errno::ECONNRESET, Errno::EPIPE, Net::SSH::Disconnect ].freeze
 
     private
       def with_ssh
@@ -191,12 +198,12 @@ class SSHKit::Backend::Netssh
         begin
           super do |ssh|
             yield ssh
-          rescue *STALE_CONNECTION_ERRORS
-            evict_stale_session(ssh)
+          rescue *STALE_CONNECTION_ERRORS => e
+            evict_stale_session(ssh) if stale_connection_error?(e)
             raise
           end
         rescue *STALE_CONNECTION_ERRORS => e
-          raise if reconnected
+          raise if reconnected || !stale_connection_error?(e)
 
           reconnected = true
           SSHKit.config.output.warn("Reconnecting to #{host}: #{e.message}")
@@ -204,6 +211,10 @@ class SSHKit::Backend::Netssh
         end
       end
 
+      def stale_connection_error?(error)
+        !error.is_a?(Net::SSH::Timeout)
+      end
+
       # `close` waits for channel-close acknowledgements, which never come over
       # a dead socket. `shutdown!` just closes the socket, and a closed session
       # is what makes the pool drop it instead of caching it again.

data/lib/dash/version.rb

@@ -1,3 +1,3 @@
 module Dash
-  VERSION = "4.0.7"
+  VERSION = "4.0.8"
 end

metadata

@@ -1,7 +1,7 @@
 --- !ruby/object:Gem::Specification
 name: dash
 version: !ruby/object:Gem::Version
-  version: 4.0.7
+  version: 4.0.8
 platform: ruby
 authors:
 - Mikael Henriksson