vtutrinov opened a new pull request, #7697:
URL: https://github.com/apache/ozone/pull/7697

   ## What changes were proposed in this pull request?
   Clear non-relevant RATIS/THREE pipelines on datanode if the previous 
datanode state on the SCM side is equal to 'DEAD'
   
   There are a number of cases when a datanode can't send heartbeat requests to 
SCM and the latter starts to handle the datanode as a **DEAD** one: close 
pipelines, and clear the command queue for the **DEAD** node. And the datanode 
will never get the commands to clear/close its pipelines, but can get a new 
command queue to create a bunch of new pipelines. The pipelines count increases 
and each node restart triggers reading of the pipelines (aka raft group) and 
can consume a lot of time and memory.
   
   We know that in the case of the DEAD state of the node, the related 
pipelines are already closed and irrelevant, and it doesn't make sense to 
initiate the raft logs on starting/restarting the datanode, and it seems we 
could delete the directories of the pipelines/raft_logs in case of the 
previously saved state of the datanode is equal to 'DEAD'
   
   ## What is the link to the Apache JIRA
   https://issues.apache.org/jira/browse/HDDS-12080
   
   ## How was this patch tested?
   manually (test cases are in progress of development and discussable)
   


-- 
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.

To unsubscribe, e-mail: [email protected]

For queries about this service, please contact Infrastructure at:
[email protected]


---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]

Reply via email to