[
https://issues.apache.org/jira/browse/IGNITE-3479?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16114262#comment-16114262
]
Semen Boikov edited comment on IGNITE-3479 at 8/4/17 12:30 PM:
---------------------------------------------------------------
This assumes following tasks:
- implement coordinator protocol (assign tx ID, return ID to use for query, etc)
- implement coordinator assignment logic (assign automatically or using
user-provided code/configuration)
Most probably there should be special 'coordinator' role for node, so that this
node should not execute other tasks.
was (Author: sboikov):
This assumes following tasks:
- implement coordinator protocol (assign tx ID, return ID to use for query, etc)
- implement coordinator assignment logic (assign automatically or using
user-provided code/configuration)
> Coordinator(s) for global transaction ordering
> ----------------------------------------------
>
> Key: IGNITE-3479
> URL: https://issues.apache.org/jira/browse/IGNITE-3479
> Project: Ignite
> Issue Type: Sub-task
> Components: cache
> Reporter: Alexey Goncharuk
> Assignee: Semen Boikov
> Fix For: 2.2
>
>
> Current transaction ID will not work for SQL MVCC ordering because two
> transaction IDs are not in total order across the cluster.
> One of the approaches is to have a dedicated coordinator which will assign a
> continuously growing transaction ID for each committed transaction. This ID
> must be acquired by each transaction at the last step of PREPARE stage.
--
This message was sent by Atlassian JIRA
(v6.4.14#64029)
