[
https://issues.apache.org/jira/browse/IOTDB-1136?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17274179#comment-17274179
]
Xiangdong Huang commented on IOTDB-1136:
----------------------------------------
Let's discuss it before repairing it..
1. For "Regular" encoding method. I do not think it should rollback the WAL if
the data is not in a regular frequency. The intention of the encoding method
is that if the data is in regular frequency, we encode the data directly. If
not, we will fill some points to make the data "regular" (The worst case is,
the timestamp gap between two adjacent points = 1).
2.
> IoTDBDescriptor.getInstance().getConfig().setReadOnly(true), which will lead
> to MultiFileLogNodeManager#forceTask useless.
I agree to trying the best to flush the WAL data to the disk. But we have to
know sometimes a flush error means no disk space left or the disk is broken
(e.g., unmounted).
> Improved reliability in flush error
> -----------------------------------
>
> Key: IOTDB-1136
> URL: https://issues.apache.org/jira/browse/IOTDB-1136
> Project: Apache IoTDB
> Issue Type: Improvement
> Components: Core/Engine
> Reporter: Chao Wang
> Assignee: Chao Wang
> Priority: Critical
>
> current, when flush data error ,
> IoTDBDescriptor.getInstance().getConfig().setReadOnly(true), which will lead
> to MultiFileLogNodeManager#forceTask useless.
> But do not have a mechanism， which recover forceTask... which this will lead
> to write failed....
> so I will do this recover... Automatic fault recovery.
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
