Run ssh with ServerAliveInterval 60

So that stalled transfers will be noticed within about 3 minutes,
even if TCPKeepAlive is disabled or doesn't work.

Rather than setting with -o, use -F with another config file,
so that any settings in ~/.ssh/config or /etc/ssh/ssh_config overrides this.

Annex/Locations.hs

@@ -63,6 +63,7 @@ module Annex.Locations (
 	gitAnnexUrlFile,
 	gitAnnexTmpCfgFile,
 	gitAnnexSshDir,
+	gitAnnexSshConfig,
 	gitAnnexRemotesDir,
 	gitAnnexAssistantDefaultDir,
 	HashLevels(..),
@@ -402,6 +403,10 @@ gitAnnexTmpCfgFile r = gitAnnexDir r </> "config.tmp"
 gitAnnexSshDir :: Git.Repo -> FilePath
 gitAnnexSshDir r = addTrailingPathSeparator $ gitAnnexDir r </> "ssh"
 
+{- .git/annex/ssh.config is used to configure ssh. -}
+gitAnnexSshConfig :: Git.Repo -> FilePath
+gitAnnexSshConfig r = gitAnnexDir r </> "ssh.config"
+
 {- .git/annex/remotes/ is used for remote-specific state. -}
 gitAnnexRemotesDir :: Git.Repo -> FilePath
 gitAnnexRemotesDir r = addTrailingPathSeparator $ gitAnnexDir r </> "remotes"

Annex/Ssh.hs

@@ -33,6 +33,7 @@ import qualified Git.Url
 import Config
 import Annex.Path
 import Utility.Env
+import Utility.Tmp
 import Types.CleanupActions
 import Git.Env
 #ifndef mingw32_HOST_OS
@@ -49,13 +50,33 @@ sshOptions (host, port) gc opts = go =<< sshCachingInfo (host, port)
 	go (Just socketfile, params) = do
 		prepSocket socketfile
 		ret params
-	ret ps = return $ concat
+	ret ps = do
-		[ ps
+		overideconfigfile <- fromRepo gitAnnexSshConfig
-		, map Param (remoteAnnexSshOptions gc)
+		-- We assume that the file content does not change.
-		, opts
+		-- If it did, a more expensive test would be needed.
-		, portParams port
+		liftIO $ unlessM (doesFileExist overideconfigfile) $
-		, [Param "-T"]
+			viaTmp writeFile overideconfigfile $ unlines
-		]
+				-- ssh expands "~"
+				[ "Include ~/.ssh/config"
+				-- ssh will silently skip the file
+				-- if it does not exist
+				, "Include /etc/ssh/ssh_config"
+				-- Everything below this point is only
+				-- used if there's no setting for it in
+				-- the above files.
+				--
+				-- Make sure that ssh detects stalled
+				-- connections.
+				, "ServerAliveInterval 60"
+				]
+		return $ concat
+			[ ps
+			, [Param "-F", File overideconfigfile]
+			, map Param (remoteAnnexSshOptions gc)
+			, opts
+			, portParams port
+			, [Param "-T"]
+			]
 
 {- Returns a filename to use for a ssh connection caching socket, and
  - parameters to enable ssh connection caching. -}

CHANGELOG

@@ -14,6 +14,10 @@ git-annex (6.20161013) UNRELEASED; urgency=medium
     NFS sometimes puts in a directory when a file is being deleted.
   * If a transfer fails for some reason, but some data managed to be sent,
     the transfer will be retried. (The assistant already did this.)
+  * Run ssh with ServerAliveInterval 60, so that stalled transfers will
+    be noticed within about 3 minutes.
+    (Any setting in your ~/.ssh/config or /etc/ssh/ssh_config 
+    overrides this.)
 
  -- Joey Hess <id@joeyh.name>  Mon, 17 Oct 2016 12:46:54 -0400
 

doc/bugs/Allow_automatic_retry_git_annex_get.mdwn

@@ -59,3 +59,5 @@ SHA256E-s41311329--69c3b054a3fe9676133605730d85b7fcef8696f6782d402a524e41b836253
 
 
 [[!meta title="Detect stalled transfer and retry or abort it"]]
+
+> [[done]] --[[Joey]]

doc/bugs/Allow_automatic_retry_git_annex_get/comment_3_8b54b726c7d89587e62b6281c9cd915a._comment

@@ -6,12 +6,12 @@
 On the ssh config, one way to do it is to pass -F with a config
 file that git-annex generates. It could look like:
 
-	Include ~/.ssh/ssh.config
+	Include ~/.ssh/config
-	Include /etc/ssh/ssh.config
+	Include /etc/ssh/ssh_config
 	ServerAliveInterval 60
 
-Since ssh uses the first config setting it sees, if ~/.ssh/ssh.config
+Since ssh uses the first config setting it sees, if `~/.ssh/config`
-or /etc/ssh/ssh.config set a ServerAliveInterval that one will be used,
+or `/etc/ssh/ssh_config` set a ServerAliveInterval that one will be used,
 and otherwise the value git-annex sets will be used.
 
 But.. Ssh enables TCPKeepAlive by default. You'd think that would be enough
@@ -23,4 +23,6 @@ or a firewall does not support it.
 
 If the problem is that users are disabling TCPKeepAlive, then
 having git-annex enable ServerAliveInterval makes sense.
+
+Ok; implemented this.
 """]]