[
https://issues.apache.org/jira/browse/FLINK-4348?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15535299#comment-15535299
]
ASF GitHub Bot commented on FLINK-4348:
---------------------------------------
Github user KurtYoung commented on the issue:
https://github.com/apache/flink/pull/2571
@mxm Thanks for the simplification, i like the idea. When i wrote the first
version of the SlotManager, i have noticed maybe i made things too complicated,
but i didn't figure out how to make things more simplify.
As it turns out, your modification covered two main problems i have faced:
1. What information to exchange during heartbeats, and what actions should
we take
2. What action should we take when the allocation failed at TaskManager
But what i really want to find out is: In this case, is there a simple
paradigm which we can follow to make whole thing clear and robust. What i
previous choose is: Take actions based on my newest runtime information. But as
you can see, it leads me to a very complex solution, each time i decide what
action should be taken, i should to check all related information and consider
all possibilities. (even it seems hard to understand why that will happen).
Your modification gives me some tips, maybe we can simplify it with
following ways:
1. RM and TM only exchange information when needed ( so heartbeat dont sync
status )
2. TM only report informations which it can changed by itself ( like slot
be free again )
Here is some thoughts about the modification:
1. We can remove the update status part entirely, since it can only do new
slot registration now, we can just move it to the task executor first
registration.
2. Once a slot becomes free in TM, notify RM
3. TM should attach the slot usage when rejecting the allocation from RM
Here is some minor problems i found in this modification:
a. As beyond1920 metioned, we dont have a way to find out a slot becomes
free ( this can be done by 2)
b. When we handleSlotRequestFailedAtTaskManager, we will make this slot
free again. If the slot is occupied by some other task now, we will
continuously failed for all allocation on this slot. ( this can be fixed by 3)
> Implement slot allocation protocol with TaskExecutor
> ----------------------------------------------------
>
> Key: FLINK-4348
> URL: https://issues.apache.org/jira/browse/FLINK-4348
> Project: Flink
> Issue Type: Sub-task
> Components: Cluster Management
> Reporter: Kurt Young
> Assignee: Maximilian Michels
>
> When slotManager finds a proper slot in the free pool for a slot request,
> slotManager marks the slot as occupied, then tells the taskExecutor to give
> the slot to the specified JobMaster.
> when a slot request is sent to taskExecutor, it should contain following
> parameters: AllocationID, JobID, slotID, resourceManagerLeaderSessionID.
> There exists 3 following possibilities of the response from taskExecutor, we
> will discuss when each possibility happens and how to handle.
> 1. Ack request which means the taskExecutor gives the slot to the specified
> jobMaster as expected.
> 2. Decline request if the slot is already occupied by other AllocationID.
> 3. Timeout which could caused by lost of request message or response message
> or slow network transfer.
> On the first occasion, ResourceManager need to do nothing. However, under the
> second and third occasion, ResourceManager need to notify slotManager,
> slotManager will verify and clear all the previous allocate information for
> this slot request firstly, then try to find a proper slot for the slot
> request again. This may cause some duplicate allocation, e.g. the slot
> request to TaskManager is successful but the response is lost somehow, so we
> may request a slot in another TaskManager, this causes two slots assigned to
> one request, but it can be taken care of by rejecting registration at
> JobMaster.
> There are still some question need to discuss in a step further.
> 1. Who send slotRequest to taskExecutor, SlotManager or ResourceManager? I
> think it's better that SlotManager delegates the rpc call to ResourceManager
> when SlotManager need to communicate with outside world. ResourceManager
> know which taskExecutor to send the request based on ResourceID. Besides this
> RPC call which used to request slot to taskExecutor should not be a
> RpcMethod, because we hope only SlotManager has permission to call the
> method, but the other component, for example JobMaster and TaskExecutor,
> cannot call this method directly.
> 2. If JobMaster reject the slot offer from a TaskExecutor, the TaskExecutor
> should notify the free slot to ResourceManager immediately, or wait for next
> heartbeat sync. The advantage of first way is the resourceManager’s view
> could be updated faster. The advantage of second way is save a RPC method in
> ResourceManager.
> 3. There are two communication type. First, the slot request could be sent as
> an ask operation where the response is returned as a future. Second,
> resourceManager send the slot request in fire and forget way, the response
> could be returned by an RPC call. I prefer the first one because it is more
> simple and could save a RPC method in ResourceManager (for callback in the
> second way).
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
