[
https://issues.apache.org/jira/browse/FLINK-15959?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17041609#comment-17041609
]
Yangze Guo commented on FLINK-15959:
------------------------------------
Hi, [~liuyufei]. I have basic some ideas from the perspective of SlotManager to
share. Hope these could help you.
>From the perspective of SlotManager, the min/max limitation is mainly related
>to the "slots" and "pendingSlots" of SlotManager. Thus, I want to first
>summarize the scenarios in which we maintain these two collections:
*Regarding the "slots"*:
* The number of elements will be reduced when TaskExecutor closed because of
heartbeat timeout, idle timeout, etc.
* The number of elements will be increased only in the registration of new
TaskExecutor.
*Regarding the "pendingSlots"*:
* The number of elements will be reduced only in the registration of new
TaskExecutor.
* The number of elements will be increased only in the resource allocation.
The union of these two collections defines the total resource/slots the
SlotManager expect, especially after the FLINK-14106. So, both minimum/maximum
limitations should act on the total number of slots of both existing and
pending.
*Regarding the min limit*, the sum of the two collections would reduce only
when TaskExecutor closed. We need the check the min limit and decide should we
allocate new resources in that scenario.
*Regarding the max limit*,
* We need to check the max limit before we allocate resource in
SlotManager#registerSlot
* When TaskExecutor closed, we also need to check if there is pending
SlotReuqest which not assign to a slot or pending slot. If there is, we need to
allocate resources while respecting the max limit.
> Add min/max number of slots configuration to limit total number of slots
> ------------------------------------------------------------------------
>
> Key: FLINK-15959
> URL: https://issues.apache.org/jira/browse/FLINK-15959
> Project: Flink
> Issue Type: Improvement
> Components: Runtime / Coordination
> Affects Versions: 1.11.0
> Reporter: YufeiLiu
> Priority: Major
>
> Flink removed `-n` option after FLIP-6, change to ResourceManager start a new
> worker when required. But I think maintain a certain amount of slots is
> necessary. These workers will start immediately when ResourceManager starts
> and would not release even if all slots are free.
> Here are some resons:
> # Users actually know how many resources are needed when run a single job,
> initialize all workers when cluster starts can speed up startup process.
> # Job schedule in topology order, next operator won't schedule until prior
> execution slot allocated. The TaskExecutors will start in several batchs in
> some cases, it might slow down the startup speed.
> # Flink support
> [FLINK-12122|https://issues.apache.org/jira/browse/FLINK-12122] [Spread out
> tasks evenly across all available registered TaskManagers], but it will only
> effect if all TMs are registered. Start all TMs at begining can slove this
> problem.
> *suggestion:*
> * Add config "taskmanager.minimum.numberOfTotalSlots" and
> "taskmanager.maximum.numberOfTotalSlots", default behavior is still like
> before.
> * Start plenty number of workers to satisfy minimum slots when
> ResourceManager accept leadership(subtract recovered workers).
> * Don't comlete slot request until minimum number of slots are registered,
> and throw exeception when exceed maximum.
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
