Layer 2 allows you to create one process per one event. However, as with default rsync behavior you might want to call one process for several events. This can be done with inlets. When any event becomes ready Lsyncd calls the action entry with inlet as parameter. The inlet can be used to grab ready single events or event lists.
For example this is the action used by default.rsync:
action = function( inlet )
local elist = inlet.getEvents( )
local config = inlet.getConfig( )
local paths = elist.getPaths( )
log( "Normal", "rsyncing list\n", table.concat( paths, '\n' ) )
spawn(elist, '/usr/bin/rsync',
'<', table.concat( paths, '\000' ),
'--delete',
config.rsync._computed,
'--from0',
'--include-from=-',
'--exclude=*',
config.source,
config.target
)
endInlet functions are:
| Function | Description |
|---|---|
| inlet.getEvent() | Retrieves the next event as in Layer 2 configuration. Multiple calls to getEvent() will return the same event unless it has spawn{}ed an action. |
| inlet.getEvents(test) | Returns a list of all events that are ready. test is optional for a function that will be called for every event to test if it should be included in the list. It has one parameter the event and returns true if an event should be included. If nil every ready event will be included in the list |
| inlet.discardEvent() | Discards an event. The next call to getEvent will thus receive another event, even if no action has been spawned for this event |
| inlet.getConfig() | returns the same as event.config. The configuration of the sync{} |
| inlet.addExclude() | adds an exclusion pattern to this sync (see Exclusions) |
| inlet.rmExclude() | removes an exclusion pattern from this sync |
| inlet.createBlanketEvent() | puts an event on the top of the Delay FIFO that blocks all events and is blocked by all events. This is used for onStartup. |
The list returned by getEvents can be handed to spawn{} as agent just as well as singular events.
Lists have following functions
| Function | Description |
|---|---|
| elist.getPaths(delimiter) | returns a string of the paths (as in event.path separated by delimiter. By default \n is used as delimiter. |
| elist.getSourcePaths(delimiter) | returns a string of the sourcePaths (as in event.sourcePath separated by delimiter. By default \n is used as delimiter. |
Take care calling getEvents() and its function since depending on the amount of events, they will cause quite some CPU load.
Layer 2 functions is nothing else than following layer 1 action loaded by the default if the user script did not provide one itself.
-----
-- Default action calls user scripts on**** functions.
--
action = function( inlet )
-- in case of moves getEvent returns the origin and destination of the move
local event, event2 = inlet.getEvent( )
local config = inlet.getConfig( )
local func = config[ 'on'.. event.etype ]
if func
then
func( event, event2 )
end
-- if function didnt change the wait status its not interested
-- in this event -> drop it.
if event.status == "wait"
then
inlet.discardEvent( event )
end
end,Lsyncd will automatically split Move events into Create and Delete events if no “onMove” field is found in the config. When handling moves in layer 1 action function, simply set “onMove” to be “true”.
Other than action Lsyncd calls init for each sync{} on initialization. This is the default init function which is loaded if the user script does not have one. It provides the onStartup() functionality for layer 2 and 3.
-----
-- called on (re)initalizing of lsyncd.
--
init = function( inlet )
local config = inlet.getConfig( )
-- calls a startup if provided by user script.
if type( config.onStartup ) == "function"
then
local event = inlet.createBlanketEvent( )
config.onStartup( event )
if event.status == 'wait'
then
-- user script did not spawn anything
-- thus the blanket event is deleted again.
inlet.discardEvent( event )
end
end
end,As another example this is the init of default.rsync. As specialty it changes the configuration in that it adds a slash to target if not there already.
-----
-- Spawns the recursive startup sync
--
init = function( inlet )
local config = inlet.getConfig( )
local event = inlet.createBlanketEvent( )
if string.sub(config.target, -1) ~= "/"
then
config.target = config.target .. "/"
end
log("Normal", "recursive startup rsync: ", config.source,
" -> ", config.target)
spawn(event,
"/usr/bin/rsync",
"--delete",
config.rsync._computed .. "r",
config.source,
config.target
)
end,When child processes are finished and their zombie processes are collected, Lsyncd calls the function of the collect entry. When collect return “again” the status of the agent (an event or an event list) will be set on “wait” again, and will become ready in delay seconds (or 1 second if smaller).
The default collect function looks in the exitcodes[] table for an entry of the exit code. Otherwise most of the unfortunately longer code below does nothing but making nice log message.
-----
-- Called when collecting a finished child process
--
collect = function(agent, exitcode)
local config = agent.config
if not agent.isList and agent.etype == "Blanket" then
if exitcode == 0
then
log("Normal", "Startup of '",agent.source,"' finished.")
elseif config.exitcodes and
config.exitcodes[exitcode] == "again"
then
log("Normal",
"Retrying startup of '",agent.source,"'.")
return "again"
else
log("Error", "Failure on startup of '",agent.source,"'.")
terminate(-1) -- ERRNO
end
return
end
local rc = config.exitcodes and config.exitcodes[exitcode]
if rc == "die"
then
return rc
end
if agent.isList
then
if rc == "again"
then
log("Normal", "Retrying a list on exitcode = ",exitcode)
else
log("Normal", "Finished a list = ",exitcode)
end
else
if rc == "again"
then
log("Normal", "Retrying ",agent.etype,
" on ",agent.sourcePath," = ",exitcode)
else
log("Normal", "Finished ",agent.etype,
" on ",agent.sourcePath," = ",exitcode)
end
end
return rc
end,