Todd Lipcon has posted comments on this change. ( http://gerrit.cloudera.org:8080/12389 )
Change subject: WIP: add release notes for 1.9.0 ...................................................................... Patch Set 1: (14 comments) http://gerrit.cloudera.org:8080/#/c/12389/1/docs/release_notes.adoc File docs/release_notes.adoc: http://gerrit.cloudera.org:8080/#/c/12389/1/docs/release_notes.adoc@51 PS1, Line 51: * KuduTable::ListPartitions() is now explicitly marked as private API. Looking at this, it looks like it was always unusable by clients because the class 'Partition' is not exported as part of the client library, right? ie a user could never have called this method to begin with because they couldn't make a 'vector<Partition>'. Given that, does it need to be release-noted at all? http://gerrit.cloudera.org:8080/#/c/12389/1/docs/release_notes.adoc@52 PS1, Line 52: * Removed private API methods from the auto-generated documentation for Kudu : C++ client. > I'm not sure that's a deprecation, this is just an improvement on the docs +1 http://gerrit.cloudera.org:8080/#/c/12389/1/docs/release_notes.adoc@62 PS1, Line 62: administrative : documentation we should link to the appropriate docs here, and maybe even remove some of the specific flag information from this release note to make it more concise http://gerrit.cloudera.org:8080/#/c/12389/1/docs/release_notes.adoc@65 PS1, Line 65: subdirectory subdirectory of the source repository? http://gerrit.cloudera.org:8080/#/c/12389/1/docs/release_notes.adoc@66 PS1, Line 66: official artifactory has been created for Apache Kudu Docker artifacts is "artifactory" the right term for docker? I thought it was a repository http://gerrit.cloudera.org:8080/#/c/12389/1/docs/release_notes.adoc@72 PS1, Line 72: * Table creation checks now consider the desired number of replicas instead of : the number of tablets. The default maximum size of a new table has effectively : been cut by a factor of the table's replication factor. writing nit: "Table creation checks now consider" was hard to parse because of "checks" being either a verb or a noun. I think it would be better to say "When creating a table, the master now enforces a restriction on the total number of replicas rather than the total number of partitions." That said, I thought we also bumped up the default for this value so effectively there is no change here for people using 3 replicas, and an increase for people using 1 replica? http://gerrit.cloudera.org:8080/#/c/12389/1/docs/release_notes.adoc@79 PS1, Line 79: * A tablet-level metric is added to indicate how much a replica needs to be can we say what metric this is? http://gerrit.cloudera.org:8080/#/c/12389/1/docs/release_notes.adoc@80 PS1, Line 80: compacted, as indicated by the average number of rowsets per unit of keyspace. is this documented anywhere in our troubleshooting docs or somesuch? http://gerrit.cloudera.org:8080/#/c/12389/1/docs/release_notes.adoc@83 PS1, Line 83: Multi-column scans of UPDATE heavy tables nit: I think this would be easier to understand if you said: "Scans which read multiple columns of tables undergoing a heavy UPDATE workload..." May also be good to say something about the relative performance improvement possible. maybe "In some cases, scan performance of such tables may be several times faster upon upgrading to this release." http://gerrit.cloudera.org:8080/#/c/12389/1/docs/release_notes.adoc@88 PS1, Line 88: KUDU-2636). This makes me raise my eyebrows a bit. It sounds like a good thing. If it's a good thing why do I need to set a config? Also it seems this flag is marked experimental, and best I can tell from looking at the git log, we defaulted it to false because we found it made some tests crashy? Given that, maybe we should just wait until our next release, make it on by default, and document it as an improvement then? http://gerrit.cloudera.org:8080/#/c/12389/1/docs/release_notes.adoc@91 PS1, Line 91: * Apache Spark was upgraded to 2.4.0. As a part of this upgrade `spark-avro` is this an incomptiable change if I'm using Kudu with an earlier version of Spark? Maybe it should be moved up to the incompatible changes section and explicitly stated whether you can use kudu-spark from this kudu release against an older version of spark? http://gerrit.cloudera.org:8080/#/c/12389/1/docs/release_notes.adoc@98 PS1, Line 98: Spark integration documentation here has been updated to reflect this yea I think in general linking the word 'here' is also bad style http://gerrit.cloudera.org:8080/#/c/12389/1/docs/release_notes.adoc@108 PS1, Line 108: * Fixed an crash caused by a race between altering tablet schemas and deleting typo: a crash http://gerrit.cloudera.org:8080/#/c/12389/1/docs/release_notes.adoc@167 PS1, Line 167: form typo -- To view, visit http://gerrit.cloudera.org:8080/12389 To unsubscribe, visit http://gerrit.cloudera.org:8080/settings Gerrit-Project: kudu Gerrit-Branch: branch-1.9.x Gerrit-MessageType: comment Gerrit-Change-Id: I733dfae39c06f15f7f55ae823678caf6ca433bfc Gerrit-Change-Number: 12389 Gerrit-PatchSet: 1 Gerrit-Owner: Andrew Wong <aw...@cloudera.com> Gerrit-Reviewer: Alexey Serbin <aser...@cloudera.com> Gerrit-Reviewer: Andrew Wong <aw...@cloudera.com> Gerrit-Reviewer: Kudu Jenkins (120) Gerrit-Reviewer: Todd Lipcon <t...@apache.org> Gerrit-Comment-Date: Sat, 09 Feb 2019 00:28:57 +0000 Gerrit-HasComments: Yes