>>> Survey results (Marton)
* 22 answers (4 without time zones)
* A separated mail will be sent out with detailed summary (and
detailed results)
* TLDR: we can introduce a new time (alternating with the old one)
which is EMEA/APAC friendly
>>> Quick status
* During a few questions about the current state, it was asked if the
viewfs change (use Ozone without changing the defaultFs) is compatible
with Hadoop 2.x line
* Uma: No technical restrictions. If it's required, can be backported.
>>> DN Audit log refactor (Dinesh)
* Audit log the chunk writes is a big problem for performance
* https://issues.apache.org/jira/browse/HDDS-3694
* Jitendra: sometimes we use audit log as a debug log :-(
* Marton: 'ozone insight' can help to get DEBUG level logs easily
* Xiaoyu: Is the audit log is async? Dinesh: it uses log4j2, yes,
it's async by default
>>> Volume quote is not enforced? (Dinesh)
* Mukul: checking is not implemented
* Question Marton: should be decided what is the plan for next
release/GA. We should either implement it or hide/remove it from the
Client API.
>>> Ratis/TWO replication (Xiaoyu/...)
* Xiaoyu summarized the current state based on offline syncs with
Tencent devs
* pipeline TWO is hard, [OPEN: Ratis/THREE --> CLOSED: TWO] can be a
possible option
* But there are some open questions regarding to the transactions
(BCSID?)
* Sid: Can Erasure Coding (EC) help instead of TWO?
* Jitendra: HDFS supports it's but didn't see much adoptions. Would
be great to understand the business requirement.
* ACTION ITEM: Start a discussion with the initial thoughts about the
EC (Marton, Uma)
>>> Backward compatibility (Arpit, Marton)
* Hadoop has a well defined compatibility guideline
* There were (rare) exceptions but good to have it as a baseline
*
https://hadoop.apache.org/docs/stable/hadoop-project-dist/hadoop-common/AdminCompatibilityGuide.html
* Would be great to adopt it and modify it to clearly communicate
what can be expected
* Current compatibility practices are driven by the demand from the
early adopters
* But would be great to write it down
>>> Rook Ozone support (Marton)
* Rook is CNCF project, a generic operator for multiple storage
* Design doc is added to extend it for Ozone
* PR is pending since a long time
* We need volunteers to help with this effort as the original
contributor (Marton) has minimal time
* If you interested: Ping Marton or help with
https://github.com/rook/rook/pull/4013
>>> Release 0.6.0 (Nanda, Sammi)
* Jitendra: would be great to have "date" defined releases as Ozone
is moving fast (release what we have after 2-3 months)
* Sammi Chen will be the Release Manager
* A few missing items are identified in the Roadmap (and already
added: s3 volume mapping, hadoop 2.x classpath changes, recon)
* Two big features which would be great to merge
* ofs (vote is required to merge, can be started in the next ~2 weeks)
* s3 volumes / bind mounting (required for better backward
compatibility)
* https://cwiki.apache.org/confluence/display/HADOOP/Ozone+Road+Map
* Follow #ozone-denali-release on the Apache slack
Marton
ps: Hope I didn't miss anything. Please reply if you think so....
---------------------------------------------------------------------
To unsubscribe, e-mail: ozone-dev-unsubscr...@hadoop.apache.org
For additional commands, e-mail: ozone-dev-h...@hadoop.apache.org
