[jira] [Commented] (ASTERIXDB-1045) Log analysis after crash may not handle logs with truncated or corrupt records correctly

["Ian Maxon (JIRA)" <ji...@apache.org>]

    [ https://issues.apache.org/jira/browse/ASTERIXDB-1045?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14907424#comment-14907424 ] 

Ian Maxon commented on ASTERIXDB-1045:
--------------------------------------

After quite a bit more discussion, the above will be done, but the IO method should be altered as well. Right now we basically fill a buffer with data from the log file, and parse it until the buffer underflows, and then fill the buffer once more by catching the exception. This could definitely be improved in a number of ways, but the tricky bit is that  the log records are not of a fixed size and do not contain a length at the beginning. One can only determine the boundary by trying to read into the log record.

> Log analysis after crash may not handle logs with truncated or corrupt records correctly
> ----------------------------------------------------------------------------------------
>
>                 Key: ASTERIXDB-1045
>                 URL: https://issues.apache.org/jira/browse/ASTERIXDB-1045
>             Project: Apache AsterixDB
>          Issue Type: Bug
>          Components: AsterixDB, Transactions
>            Reporter: asterixdb-importer
>            Assignee: Ian Maxon
>            Priority: Critical
>
> Log analysis after crash may not handle logs with truncated or corrupt records correctly



--
This message was sent by Atlassian JIRA
(v6.3.4#6332)