[jira] [Updated] (IGNITE-18181) Race during table creation

["Roman Puchkovskiy (Jira)" <ji...@apache.org>]

     [ https://issues.apache.org/jira/browse/IGNITE-18181?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]

Roman Puchkovskiy updated IGNITE-18181:
---------------------------------------
    Description: 
{{tablesByIdVV.update()}} inside {{TableManager#updateAssignments}} initiates async RAFT groups' start, but its future supplier returns {{{}completedFuture(null){}}}, so there is no causal dependency between completion of what {{update()}} returned and the futures that start RAFT groups; hence, there is no causal dependency between completing 'table created' future and RAFT groups start.

While {{updateAssignmentsInternal}} is fully synchronous, this is not a problem. But in IGNITE-18170 we make it async, and here the race starts manifesting itself.

Probably, update supplier should return a future depending on RAFT group start future and not just a completed future.

  was:TBD


> Race during table creation
> --------------------------
>
>                 Key: IGNITE-18181
>                 URL: https://issues.apache.org/jira/browse/IGNITE-18181
>             Project: Ignite
>          Issue Type: Bug
>            Reporter: Roman Puchkovskiy
>            Assignee: Roman Puchkovskiy
>            Priority: Major
>              Labels: ignite-3
>             Fix For: 3.0.0-beta2
>
>          Time Spent: 10m
>  Remaining Estimate: 0h
>
> {{tablesByIdVV.update()}} inside {{TableManager#updateAssignments}} initiates async RAFT groups' start, but its future supplier returns {{{}completedFuture(null){}}}, so there is no causal dependency between completion of what {{update()}} returned and the futures that start RAFT groups; hence, there is no causal dependency between completing 'table created' future and RAFT groups start.
> While {{updateAssignmentsInternal}} is fully synchronous, this is not a problem. But in IGNITE-18170 we make it async, and here the race starts manifesting itself.
> Probably, update supplier should return a future depending on RAFT group start future and not just a completed future.



--
This message was sent by Atlassian Jira
(v8.20.10#820010)