[ 
https://issues.apache.org/jira/browse/FLUME-1527?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
 ]

Brock Noland updated FLUME-1527:
--------------------------------

    Description: 
At startup we seed the sequence numbers for transaction id and write order id 
with the highest of the following:

-current time in ms
-the highest id we find from the log files we find during replay

There is a small and unlikely hole in this logic. If an agent had received tons 
of events and thus the sequence numbers had incremented past current time in 
ms, then checkpointed, received no more events (as such nothing is replayed), 
and restarted we would not have any logs to replay and as such would see the 
sequence numbers with only current time in ms.

  was:
At startup we seed the sequence numbers for transaction id and write order id 
with the highest of the following:

-current time in ms
-the highest id we find from the log files

There is a small and unlikely hole in this logic. If an agent had received tons 
of events and thus the sequence numbers had incremented past current time in 
ms, then checkpointed, received no more events, and restarted we would not have 
any logs to replay and as such would see the sequence numbers with only current 
time in ms.

    
> Seed TransactionID/WriteOrderID in the case no logs need to be replayed
> -----------------------------------------------------------------------
>
>                 Key: FLUME-1527
>                 URL: https://issues.apache.org/jira/browse/FLUME-1527
>             Project: Flume
>          Issue Type: Bug
>          Components: Channel
>            Reporter: Brock Noland
>
> At startup we seed the sequence numbers for transaction id and write order id 
> with the highest of the following:
> -current time in ms
> -the highest id we find from the log files we find during replay
> There is a small and unlikely hole in this logic. If an agent had received 
> tons of events and thus the sequence numbers had incremented past current 
> time in ms, then checkpointed, received no more events (as such nothing is 
> replayed), and restarted we would not have any logs to replay and as such 
> would see the sequence numbers with only current time in ms.

--
This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators
For more information on JIRA, see: http://www.atlassian.com/software/jira

Reply via email to