hadoop-hdfs-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Hongxu Ma <inte...@outlook.com>
Subject Re: About HDFS consistency model
Date Wed, 02 Aug 2017 03:18:26 GMT
Thanks again, Ravi!

These links are helpful, a good root for me to trace this topic, maybe I can find more useful
info, thanks!

在 01/08/2017 04:09, Ravi Prakash 写道:
Hi Hongxu!

It might be best to go through JIRA comments. Some of them have design docs too. e.g. https://issues.apache.org/jira/browse/HDFS-265
, https://issues.apache.org/jira/browse/HDFS-744 etc.

I know there may be some windy conversations, but the good thing about open-source is that
all these deliberations are for everyone to see.


On Mon, Jul 31, 2017 at 3:13 AM, Hongxu Ma <interma@outlook.com<mailto:interma@outlook.com>>
Hi Dev
In short, is there a full introduction to hdfs consistency model?

I have already read some miscellaneous stuffs, e.g.
- some simple scenario
    - read+read, ok
    - write+write, forbidden (guaranteed be lease)
- design doc of append, truncate
- hflush/hsync
- top pages of google "hdfs consistency model"

But I didn't find a detailed doc/link to clarify this topic, especially how to deal with concurrent
read+write, e.g.:
Can a later reader can read the contents which the writer (doesn't close the file yet) appended?
Is hflush/hsync the only ways to ensure the consistency? etc..

I known the GFS paper has a specific section to clarify this topic, so I think hdfs may has
a similar one.
Very thanks for helpful link.


  • Unnamed multipart/alternative (inline, None, 0 bytes)
View raw message