[
https://issues.apache.org/jira/browse/USERGRID-60?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Todd Nine updated USERGRID-60:
------------------------------
Description:
In version 2.0, we need to ensure part of the commit is that the UUID of the
entity is persistence in a timeout q. This can be done in a simple time series
(possibly just using SQS) so that if a post process fails after commit, it can
be restarted. Note that this implies that all post processing is idempotent.
This should be part of the 2.0 code base.
It should have the following properties
1) Completely persistent
2) Can execute on any node. By default, this process will happen in memory on
the node that accepts the commit. This only really applies during a
shutdown/failure scenario
3) Should provide some partitioning to distribute load among remaining nodes.
was:
In version 2.0, we need to ensure part of the commit is that the UUID of the
entity is persistence in a timeout q. This can be done in a simple time series
(possibly just using SQS) so that if a post process fails after commit, it can
be restarted. Note that this implies that all post processing is idempotent.
This should be part of the 2.0 code base.
> Ensure write consistency via Persistent timeout Q
> --------------------------------------------------
>
> Key: USERGRID-60
> URL: https://issues.apache.org/jira/browse/USERGRID-60
> Project: Usergrid
> Issue Type: Improvement
> Components: Stack
> Reporter: Todd Nine
>
> In version 2.0, we need to ensure part of the commit is that the UUID of the
> entity is persistence in a timeout q. This can be done in a simple time
> series (possibly just using SQS) so that if a post process fails after
> commit, it can be restarted. Note that this implies that all post processing
> is idempotent. This should be part of the 2.0 code base.
> It should have the following properties
> 1) Completely persistent
> 2) Can execute on any node. By default, this process will happen in memory
> on the node that accepts the commit. This only really applies during a
> shutdown/failure scenario
> 3) Should provide some partitioning to distribute load among remaining nodes.
--
This message was sent by Atlassian JIRA
(v6.1.5#6160)
