[GitHub] [hadoop-ozone] fapifta commented on pull request #1322: HDDS-3829. Introduce Layout Feature interface in Ozone.

2020-08-25 Thread GitBox


fapifta commented on pull request #1322:
URL: https://github.com/apache/hadoop-ozone/pull/1322#issuecomment-680294492


   Hi @avijayanhwx,
   
   the current test failures seems to be unrelated, as this will go into the 
feature branch and we still will work on a lot of stuff and form this further, 
I am +1 committing the changes to the branch as they are, and so we can start 
work on the initial idea further.
   Thank you for the explanation, and addressing my concerns partly here, and 
partly in other forms.
   
   As we discussed, the current aspect implementation requires the version 
manager to be static in order to be able to check versions without major 
changes in the code, we might address it later, other things seems to be 
reasonable.



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.

For queries about this service, please contact Infrastructure at:
us...@infra.apache.org



-
To unsubscribe, e-mail: ozone-issues-unsubscr...@hadoop.apache.org
For additional commands, e-mail: ozone-issues-h...@hadoop.apache.org



[GitHub] [hadoop-ozone] fapifta commented on pull request #1322: HDDS-3829. Introduce Layout Feature interface in Ozone.

2020-08-17 Thread GitBox


fapifta commented on pull request #1322:
URL: https://github.com/apache/hadoop-ozone/pull/1322#issuecomment-674766274


   Hi @avijayanhwx,
   
   thank you for starting this work, and posting an inital version of the core 
code for the upgrade support.
   I have a few general questions and concerns, also added a few comments after 
a quick review.
   
   In HDFS the layoutversion is a monotonically decreasing number, we chose to 
use monotonically increasing version numbers, I am unsure why HDFS chose the 
negative numbers, can there be some hidden considerations, we might go through, 
before committing to the positive layoutversions and monotonic increase? I 
haven't found one.
   
   Also, in HDFS one layoutversion covers one feature, would it be better to do 
not let two feetures associated with one layoutversion number? What is the 
benefit of having two features mapped to the same layout version? I don't feel 
good about it, though I don't have a specific example yet where it can cause 
trouble.
   
   LayoutVersionManager is implemented in a way that it is fully static. I am 
unsure if this is a good design, I understand the intent that there has to be 
only one instance of this per component, and seeing it this way it is a 
reasonable choice to use static a implementation, but it can fire back later 
when we want to implement tests that involves changing something in this logic, 
and we can not freely and easily change the behaviour in tests as I fear, also 
it can introduce invisible interdependencies later that might be hard to 
detect/factor out. Implementing it in a non-static way does not seem to cause 
any drawback, even we can be fine with multiple instances for the same 
component, as the used values will be anyway hardcoded in the real system. What 
do you think, I would consider making it non-static, as I think it has more 
possibilities and less limitations later.



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.

For queries about this service, please contact Infrastructure at:
us...@infra.apache.org



-
To unsubscribe, e-mail: ozone-issues-unsubscr...@hadoop.apache.org
For additional commands, e-mail: ozone-issues-h...@hadoop.apache.org