?????? [Discuss] BP-56: Support non-stop bookie data migration and bookie offline

[lordcheng10]

1.The first question, I don't understand what you mean, if it is an operation 
error, which leads to the wrong writing of the offline node, there is no way to 
avoid it.
The current offline command, if the operation is wrong, it may also be offline 
to the wrong bookie node??
2.Before going offline, we will set the bookie node to be offline to read-only 
mode, so no new data will be written to the offline node.


The current offline process is:
a. Set the bookie node to read-only mode;
b. Stop a bookie node, and then run the decommission command to trigger data 
migration;
c. Use the command bin/bookkeeper shell listledgers -bookieid <target 
bookieid&gt; to confirm whether there is still a ledger on the bookie. If there 
is no ledger on the node, the data has been migrated and the offline operation 
of the node is completed, and then you can continue Offline the next node;


Modified process:
a. Set the bookie node to read-only mode;
b. Migrate the data on the bookie node to be offline to other readable nodes by 
executing the command;
c. When the data migration is completed, stop nodes in batches;


------------------&nbsp;????????&nbsp;------------------
??????:                                                                         
                                               "dev"                            
                                                        
<lushiji2...@gmail.com&gt;;
????????:&nbsp;2022??8??25??(??????) ????6:32
??????:&nbsp;"dev"<dev@bookkeeper.apache.org&gt;;

????:&nbsp;Re: [Discuss] BP-56: Support non-stop bookie data migration and 
bookie offline



I think this feature is somewhat custom and not very generic; and there are
risks:
1.&nbsp; If you want to go offline on node A (which has already been extracted),
but wrong write B, this function will directly go offline on node B, which
is likely to cause online failures
2.&nbsp; If the node to be offline suddenly accesses traffic, how should it be
handled? It is easy to cause the loss of cluster replicas

In response to these two problems, how to avoid, please help explain

lordcheng10 <1572139...@qq.com.invalid&gt; ??2022??8??25?????? 17:12??????