add master-replica switching support

This patch adds ability to use queue in a master-replica scheme.
The queue will monitor the operation mode of the tarantool and
perform the necessary actions accordingly.

This patch adds five states for queue: INIT, STARTUP, RUNNING, ENDING
and WAITING. When the tarantool is launched for the first time,
the state of the queue is always INIT until box.info.ro is false.

States switching scheme:

                       +-----------+
                       |  RUNNING  |
                       +-----------+
                        ^         | (rw -> ro)
                        |         v
+------+       +---------+       +--------+
| INIT | --->  | STARTUP |       | ENDING |
+------+       +---------+       +--------+
                        ^         |
             (ro -> rw) |         v
                       +-----------+
                       |  WAITING  |
                       +-----------+

In the STARTUP state, the queue is waiting for possible data synchronization
with other cluster members by the time of the largest upstream lag multiplied
by two. After that, all taken tasks are released, except for tasks with
session uuid matching inactive sessions uuids. This makes possible to take
a task, switch roles on the cluster, and release the task within the timeout
specified by the queue.cfg({ttr = N}) parameter. Note: all clients that take()
and do not ack()/release() tasks must be disconnected before changing the role.
And the last step in the STARTUP state is starting tube driver using new
method called start(). Each tube driver must implement start() and stop()
methods. The start() method should start the driver fibers, if any, in other
words, initialize all asynchronous work with the tubes space. The stop()
methods shuld stop the driver fibers, if any, respectively.

In the RUNNING state, the queue is working as usually. The ENDING state calls
stop() method. in the WAITING state, the queue listens for a change in the
read_only flag.

All states except INIT is controlled by new fiber called 'queue_state_fiber'.

A new release_all() method has also been added, which forcibly returns all
taken tasks to a ready state. This method can be called per tube.

Closes #120

Author: 0x501D

queue/abstract.lua

@@ -2,8 +2,9 @@ local log      = require('log')
 local fiber    = require('fiber')
 local uuid     = require('uuid')
 
-local session  = require('queue.abstract.queue_session')
-local state    = require('queue.abstract.state')
+local session     = require('queue.abstract.queue_session')
+local state       = require('queue.abstract.state')
+local queue_state = require('queue.abstract.queue_state')
 
 local util     = require('queue.util')
 local qc       = require('queue.compat')
@@ -36,6 +37,9 @@ local queue = {
     stat = {}
 }
 
+-- Release all tasks except tasks that have session_uuid
+-- stored in inactive sessions.
+-- They will be freed later in release_session_tasks().
 local function tube_release_all_tasks(tube)
     local prefix = ('queue: [tube "%s"] '):format(tube.name)
 
@@ -51,8 +55,18 @@ local function tube_release_all_tasks(tube)
     log.info(prefix .. 'releasing all taken task (may take a while)')
     local released = 0
     for _, task in tube.raw:tasks_by_state(state.TAKEN) do
+        local taken = box.space._queue_taken_2.index.task:get{
+            tube.tube_id, task[1]
+        }
+        if taken and box.space._queue_inactive_sessions:get{taken[4]} then
+            log.info(prefix ..
+                ('skipping task: %d, tube_id: %d'):format(task[1],
+                    tube.tube_id))
+            goto next
+        end
         tube.raw:release(task[1], {})
         released = released + 1
+        ::next::
     end
     log.info(prefix .. ('released %d tasks'):format(released))
 end
@@ -73,6 +87,9 @@ end
 local tube = {}
 
 function tube.put(self, data, opts)
+    if not queue_state.check() then
+        return nil
+    end
     opts = opts or {}
     local task = self.raw:put(data, opts)
     return self.raw:normalize_task(task)
@@ -82,6 +99,9 @@ local conds = {}
 local releasing_connections = {}
 
 function tube.take(self, timeout)
+    if not queue_state.check() then
+        return nil
+    end
     timeout = util.time(timeout or util.TIMEOUT_INFINITY)
     local task = self.raw:take()
     if task ~= nil then
@@ -120,6 +140,9 @@ function tube.take(self, timeout)
 end
 
 function tube.touch(self, id, delta)
+    if not queue_state.check() then
+        return
+    end
     if delta == nil then
         return
     end
@@ -143,6 +166,9 @@ function tube.touch(self, id, delta)
 end
 
 function tube.ack(self, id)
+    if not queue_state.check() then
+        return nil
+    end
     check_task_is_taken(self.tube_id, id)
     local tube = box.space._queue:get{self.name}
     local space_name = tube[3]
@@ -169,10 +195,32 @@ local function tube_release_internal(self, id, opts, session_uuid)
 end
 
 function tube.release(self, id, opts)
+    if not queue_state.check() then
+        return nil
+    end
     return tube_release_internal(self, id, opts)
 end
 
+-- release all tasks (force release, session_id will not be used)
+function tube.release_all(self)
+    if not queue_state.check() then
+        return
+    end
+    local prefix = ('queue: [tube "%s"] '):format(self.name)
+
+    log.info(prefix .. 'releasing all taken task (may take a while)')
+    local released = 0
+    for _, task in self.raw:tasks_by_state(state.TAKEN) do
+        self.raw:release(task[1], {})
+        released = released + 1
+    end
+    log.info(prefix .. ('released %d tasks'):format(released))
+end
+
 function tube.peek(self, id)
+    if not queue_state.check() then
+        return nil
+    end
     local task = self.raw:peek(id)
     if task == nil then
         error(("Task %s not found"):format(tostring(id)))
@@ -181,6 +229,9 @@ function tube.peek(self, id)
 end
 
 function tube.bury(self, id)
+    if not queue_state.check() then
+        return nil
+    end
     local task = self:peek(id)
     local is_taken, _ = pcall(check_task_is_taken, self.tube_id, id)
     if is_taken then
@@ -193,17 +244,46 @@ function tube.bury(self, id)
 end
 
 function tube.kick(self, count)
+    if not queue_state.check() then
+        return nil
+    end
     count = count or 1
     return self.raw:kick(count)
 end
 
+function tube.start(self)
+    if not self.raw.start then
+        log.warn('queue: [tube "%s"] method start %s',
+            self.name, 'is not implemented')
+        return
+    end
+
+    return self.raw:start()
+end
+
+function tube.stop(self)
+    if not self.raw.stop then
+        log.warn('queue: [tube "%s"] method stop %s',
+            self.name, 'is not implemented')
+        return
+    end
+
+    return self.raw:stop()
+end
+
 function tube.delete(self, id)
+    if not queue_state.check() then
+        return nil
+    end
     self:peek(id)
     return self.raw:normalize_task(self.raw:delete(id))
 end
 
 -- drop tube
 function tube.drop(self)
+    if not queue_state.check() then
+        return nil
+    end
     local tube_name = self.name
 
     local tube = box.space._queue:get{tube_name}
@@ -238,6 +318,9 @@ end
 -- truncate tube
 -- (delete everything from tube)
 function tube.truncate(self)
+    if not queue_state.check() then
+        return
+    end
     self.raw:truncate()
 end
 
@@ -248,6 +331,9 @@ function tube.on_task_change(self, cb)
 end
 
 function tube.grant(self, user, args)
+    if not queue_state.check() then
+        return
+    end
     local function tube_grant_space(user, name, tp)
         box.schema.user.grant(user, tp or 'read,write', 'space', name, {
             if_not_exists = true,
@@ -296,6 +382,8 @@ local required_driver_methods = {
     'bury',
     'kick',
     'peek',
+    'stop',
+    'start',
     'touch',
     'truncate',
     'tasks_by_state'
@@ -406,6 +494,10 @@ local function release_session_tasks(session_uuid)
     end
 end
 
+function method.state()
+    return queue_state.show()
+end
+
 function method._on_consumer_disconnect()
     local conn_id = connection.id()
 
@@ -449,6 +541,9 @@ end
 -------------------------------------------------------------------------------
 -- create tube
 function method.create_tube(tube_name, tube_type, opts)
+    if not queue_state.check() then
+        return
+    end
     opts = opts or {}
     if opts.if_not_exists == nil then
         opts.if_not_exists = false
@@ -551,7 +646,7 @@ function method.start()
     if _taken == nil then
         -- tube_id, task_id, connection_id, session_uuid, time
         _taken = box.schema.create_space('_queue_taken_2', {
-            temporary = true,
+            temporary = false,
             format = {
                 {name = 'tube_id', type = num_type()},
                 {name = 'task_id', type = num_type()},
@@ -570,6 +665,10 @@ function method.start()
             parts = {4, str_type()},
             unique = false
         })
+    else
+        if _taken.temporary then
+            _taken:alter{temporary = false}
+        end
     end
 
     for _, tube_tuple in _queue:pairs() do
@@ -586,6 +685,8 @@ function method.start()
     session.start()
 
     connection.on_disconnect(queue._on_consumer_disconnect)
+    queue_state.set_running()
+    queue_state.create_state_fiber(queue.tube)
     return queue
 end
 

queue/abstract/driver/fifo.lua

@@ -148,4 +148,14 @@ function method.truncate(self)
     self.space:truncate()
 end
 
+-- this driver has no background activity
+-- implement dummy methods for the API requirement
+function method.start()
+    return
+end
+
+function method.stop()
+    return
+end
+
 return tube

queue/abstract/driver/fifottl.lua

@@ -1,4 +1,5 @@
 local log      = require('log')
+local ffi      = require('ffi')
 local fiber    = require('fiber')
 local state    = require('queue.abstract.state')
 
@@ -159,6 +160,14 @@ local function fifottl_fiber(self)
     while true do
         if box.info.ro == false then
             local stat, err = pcall(fifottl_fiber_iteration, self, processed)
+
+            if ffi.istype("struct error", err) then
+                if err.type == "FiberIsCancelled" then
+                    log.info("fifottl was cancelled")
+                    break
+                end
+            end
+
             if not stat and not (err.code == box.error.READONLY) then
                 log.error("error catched: %s", tostring(err))
                 log.error("exiting fiber '%s'", fiber.name())
@@ -365,4 +374,21 @@ function method.truncate(self)
     self.space:truncate()
 end
 
+function method.start(self)
+    if self.fiber then
+        log.info('fifottl fiber is already started')
+        return
+    end
+    self.fiber = fiber.create(fifottl_fiber, self)
+end
+
+function method.stop(self)
+    if not self.fiber then
+        log.info('fifottl fiber is not started')
+        return
+    end
+    self.fiber:cancel()
+    self.fiber = nil
+end
+
 return tube

queue/abstract/driver/utube.lua

@@ -175,4 +175,14 @@ function method.truncate(self)
     self.space:truncate()
 end
 
+-- this driver has no background activity
+-- implement dummy methods for the API requirement
+function method.start()
+    return
+end
+
+function method.stop()
+    return
+end
+
 return tube

queue/abstract/driver/utubettl.lua

@@ -1,4 +1,5 @@
 local log      = require('log')
+local ffi      = require('ffi')
 local fiber    = require('fiber')
 
 local state    = require('queue.abstract.state')
@@ -167,6 +168,14 @@ local function utubettl_fiber(self)
     while true do
         if box.info.ro == false then
             local stat, err = pcall(utubettl_fiber_iteration, self, processed)
+
+            if ffi.istype("struct error", err) then
+                if err.type == "FiberIsCancelled" then
+                    log.info("utubettl was cancelled")
+                    break
+                end
+            end
+
             if not stat and not err.code == box.error.READONLY then
                 log.error("error catched: %s", tostring(err))
                 log.error("exiting fiber '%s'", fiber.name())
@@ -392,4 +401,21 @@ function method.truncate(self)
     self.space:truncate()
 end
 
+function method.start(self)
+    if self.fiber then
+        log.info('utubettl fiber is already started')
+        return
+    end
+    self.fiber = fiber.create(utubettl_fiber, self)
+end
+
+function method.stop(self)
+    if not self.fiber then
+        log.info('utubettl fiber is not started')
+        return
+    end
+    self.fiber:cancel()
+    self.fiber = nil
+end
+
 return tube