[ 
https://issues.apache.org/jira/browse/ZOOKEEPER-4706?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17768292#comment-17768292
 ] 

Kezhu Wang edited comment on ZOOKEEPER-4706 at 9/23/23 2:06 PM:
----------------------------------------------------------------

I presented a "circumvent" to use {{zxid}} as "sequence number" with new 
{{CreateMode}}. The code is simple and demands no on-disk wire changes. But I 
feel pessimistic as it diverge sequential nodes to consecutive ones and 
non-consecutive ones. This clearly mark itself a "circumvent" but not a fix. In 
long term, I prefer to extend {{Stat.cversion}} to 64-bit as it fit ZooKeeper 
more naturally comparing to how it currently function. So, I plan to close that 
pr soon. Feel free to reopen it if any committers interest.


was (Author: kezhuw):
I presented a "circumvent" to use {{zxid}} as "sequence number" with new 
{{CreateMode}}. The code is simple and demands no on-disk wire changes. But I 
feel pessimistic as it diverge sequential nodes to consecutive ones and 
non-consecutive ones. This clearly mark itself a "circumvent" but not a fix. In 
long term, I prefer to extend {{Stat.cversion}} to 64-bit as it fit ZooKeeper 
more naturally as how it currently function. So, I plan to close that pr soon. 
Feel free to reopen it if any committers interest.

> Use 64 bits counter to ensure monotonically increasing and uniqueness of 
> sequence node name
> -------------------------------------------------------------------------------------------
>
>                 Key: ZOOKEEPER-4706
>                 URL: https://issues.apache.org/jira/browse/ZOOKEEPER-4706
>             Project: ZooKeeper
>          Issue Type: New Feature
>          Components: jute, server
>            Reporter: Li Wang
>            Priority: Major
>              Labels: pull-request-available
>             Fix For: 4.0.0
>
>          Time Spent: 10m
>  Remaining Estimate: 0h
>
> With 32 bits counter, the sequence number used for sequential node name is 
> not monotonically increase and unique any more after overflow.
> To fix the issue, we would want to use 64 bits counter.



--
This message was sent by Atlassian Jira
(v8.20.10#820010)

Reply via email to