MySQL Lists are EOL. Please join:

List:Cluster« Previous Message
From:Frazer Clement Date:February 18 2019 10:15am
Subject:Re : Incomplete log for node group
View as plain text  
> 1. I have received some error logs in data node.
> """
> *** 278 BYTES LOST ***
>    file 64(364)  FileChangeState: 0  logFileStatus: 1  currentMbyte: 255 
> currentFilepage
> 8191
>    file 35(335)  FileChangeState: 0  logFileStatus: 1  currentMbyte: 0 
> currentFilepage 0
>    file 75(375)  FileChangeState: 0  logFileStatus: 1  currentMbyte: 255 
> currentFilepage
> 8191
>    file 113(413)  FileChangeState: 0  logFileStatus: 1  currentMbyte: 255 
> currentFilepage
> 8191
>    file 76(376)  FileChangeState: 0  logFileStatus: 1  currentMbyte: 255 
> currentFilepage
> 8191
>   """
- The above looks like some info generated as part of a system self 
check failing, and the node shutting itself down.
    This happens when e.g. some critical system function, such as the 
Local or Global checkpoint mechanisms, are stalled for too long.
    The part pasted above looks like a set of open file states related 
to the Redo logs.
    However it is hard to determine more as this is just a short snippet 
- do you have more log information?  MySQL Cluster ships with a tool 
called ndb_error_reporter which can gather all the relevant logs into a 
tar.gz for uploading to a bug report on mysql.com.

- The "*** 278 BYTES LOST ***" refers to an internal circular log buffer 
overflow - so some of the diagnostic messages are lost.
>
> 2. mysql cluster restart failureļ¼š
> 	error info: Incomplete log for node group: 1!  starting nodes: 000000000000001e
This error message indicates that the nodes available for starting in 
node group 1 do not have the latest set of redo logs for that nodegroup.
This generally occurs when e.g. :

   - Nodegroup 1 contains two nodes A + B
   - Node A fails and is ejected from the cluster at time t1
   - Transactions continue processing, with Node B in Nodegroup 1
   - Later the whole cluster fails at time t2
   - The whole cluster is started, with Node A and *without* Node B
   - Nodegroup 1 (A) only has redo logs up to time t1 - it is missing 
logs between t1 and t2

Another less common scenario could be when the underlying filesystems 
are recovered from a snapshot of an older stable state.
> now i can't start my mysql cluster, How do fix my database cluster?
The solution is to recover Node B (and its Redo logs).

If this is not possible then the fallback is to restore from a Backup, 
and apply binary logs.

Hope this helps!
Frazer

Thread
Re : Incomplete log for node groupFrazer Clement18 Feb 2019