You are viewing a plain text version of this content. The canonical link for it is here.
Posted to dev@phoenix.apache.org by "Lars Hofhansl (JIRA)" <ji...@apache.org> on 2018/11/27 22:19:00 UTC
[jira] [Updated] (PHOENIX-5045) Remove Global Mutable index code
from Phoenix
[ https://issues.apache.org/jira/browse/PHOENIX-5045?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]
Lars Hofhansl updated PHOENIX-5045:
-----------------------------------
Description:
In a continued effort to reduce Phoenix' complexity I propose that we remove the current GLOBAL MUTABLE index implementation from Phoenix.
It can't be made working reliably and comes with considerable complexity such as:
* temporarily disabled indexes, or
* indexes that return stale data, and
* complex catch-up logic
* server-to-server communication with the risk of tying up handler threads
* special retry logic on the client (to avoid retries in the handler threads trying to update the index table)
IMHO it does not even matter whether we have a replacement or not. The current code should not be used.
Let's discuss.
was:
In a continued effort to reduce Phoenix' complexity I propose that we remove the current GLOBAL MUTABLE index implementation from Phoenix.
It can't be made working reliably and comes with considerable complexity such as:
* temporarily disabled indexes, or
* indexes that return stale data, and
* complex catch-up logic
* server-to-server communication with the risk of tying up handler threads
IMHO it does not even matter whether we have a replacement or not. The current code should not be used.
Let's discuss.
> Remove Global Mutable index code from Phoenix
> ---------------------------------------------
>
> Key: PHOENIX-5045
> URL: https://issues.apache.org/jira/browse/PHOENIX-5045
> Project: Phoenix
> Issue Type: Task
> Reporter: Lars Hofhansl
> Priority: Major
>
> In a continued effort to reduce Phoenix' complexity I propose that we remove the current GLOBAL MUTABLE index implementation from Phoenix.
> It can't be made working reliably and comes with considerable complexity such as:
> * temporarily disabled indexes, or
> * indexes that return stale data, and
> * complex catch-up logic
> * server-to-server communication with the risk of tying up handler threads
> * special retry logic on the client (to avoid retries in the handler threads trying to update the index table)
> IMHO it does not even matter whether we have a replacement or not. The current code should not be used.
> Let's discuss.
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)