User-facing changes in 19.2 that were not picked up in release notes #5819
Comments
|
Thanks for doing this @knz!
|
|
I have a slight dislike the framing of this and the implication that use of ‘none’ can just be assumed to be mistaken. There seems to be an assumption here that one commit = one user-facing-change, but one big monster commit is not how we usually want to go about building big new features. Many cases above include many commits working towards implementing one new user-visible feature — there is no reason for those to each have individual release notes as they are not changing an existing user-visible feature. In other cases, they are fixing bugs, but in similarly unreleased functionality and thus again, are not user-visible changes. |
|
The release notes in the PRs are meant to be accurate between alpha releases, so it is reasonable to document small changes that are part of a bigger new feature (though I agree it's not terribly useful to provide a lot of detail). How these get massaged into the release notes for a major release is not trivial - the improvements related to a bigger change do need to be coalesced into one item. |
David, I stand behind your view here. I have been careful to de-duplicate entries and collapse multiple PRs that are part of a larger whole under a single entry (for example, I removed all the "opt: implement X or Y" entries because they all fall under "remove the heuristic planner", which was documented eventually). I did the same with various IMPORT FROM PRs. Unfortunately it is hard for me to be 100% accurate so I apologize if I forgot to collapse some groups. |
|
Thank you for driving this, @knz. Since our Eng team has grown so much, I had a nagging worry that many user-facing changes weren't getting flagged for announcement and documentation. At the end of this exercise, we'll have a clear sense of that. |
|
|
|
|
* #40880 (telemetry change) Telemetry is collected when
|
|
|
Done, but I think we should have a conversation about what changes are worthy to be included in release notes. For one, I enjoy reading release notes for software, but not if they're endless. I'm afraid that activism like this, coupled with more and more monkeys banging on this, will lead to a million bullets of very varying degrees of importance. |
|
|
Thanks for picking this off! Most of these are of very marginal use to the user, but the changefeed one is a great find. The rest of mine were correctly omitted.
|
|
#40948: Release note (bug fix): Fix the outdated message if AOST is used by a query inside a transaction block. Prompts the user to use the "SET TRANSACTION AS OF SYSTEM TIME" syntax. |
|
- #39662 <cockroachdb/cockroach#39662> workload:
fall back to using 19.1 vectorize setting on syntax error. Release note
(bug fix): The querybench workload has been modified to use the correct
syntax for vectorized execution in 19.1, attempting to run this workload
with vectorized execution would previously return an error.
- #41467 <cockroachdb/cockroach#41467>
release-19.2:
colrpc: downgrade log messages from Error to Warning. Release note (bug
fix): Some log messages reported as Errors have been downgraded to Warnings.
- #37088 <cockroachdb/cockroach#37088> exec:
wrap hashRouter work with CatchVectorizedRuntimeError. Release note (bug
fix): An expected panic in the vectorized engine is now reported as an
error without crashing the server.
- #39093 <cockroachdb/cockroach#39093> exec:
wait for internal goroutine completion in UnorderedSynchronizer. Release
note (bug fix): A race condition in the vectorized engine has been fixed.
- #38006 <cockroachdb/cockroach#38006> exec:
set output columnTypes when creating noop operator. Release note (bug fix):
A bug planning noop operators in the vectorzied execution engine has been
fixed.
- #40668 <cockroachdb/cockroach#40668> exec:
reset internal unsafeBatch in orderedAggregator. Release note (bug fix): A
bug with vectorized aggregations due to incorrect state transitions has
been fixed.
- #38557 <cockroachdb/cockroach#38557> exec:
protect against unset syncFlowConsumer. Release note (bug fix): A
vectorized engine plan-time state assertion was added to avoid a panic.
- #39248 <cockroachdb/cockroach#39248> exec:
fix metadata drain cycles. Release note (bug fix): A bug that led to
queries hanging in the vectorized engine due to incorrect plans has been
fixed.
- #37571 <cockroachdb/cockroach#37571> exec:
check for unset output columnTypes. Release note (bug fix): A vectorized
engine plan-time state assertion was added.
- #38329 <cockroachdb/cockroach#38329> distsqlrun:
include StartScan stall time in EXPLAIN ANALYZE. Release note (sql change):
EXPLAIN ANALYZE now reports the time taken to look up rows in a lookup join.
- #38622 <cockroachdb/cockroach#38622> distsqlrun:
bump index joiner batch size from 100 to 10k. Release note (performance
improvement): Index join performance has been improved.
- #37745 <cockroachdb/cockroach#37745>
col{data,serde}:
various fixes related to TestArrowBatchConverterRandom (2 commits). Release
note (bug fix): A bug in data serialization in the vectorized engine caused
by incorrectly determining the presence of nulls in rows has been fixed.
…On Mon, Nov 11, 2019 at 10:48 AM irfan sharif ***@***.***> wrote:
- #41764: Release note (bug fix): Previously a disk stall could allow
a node to continue heartbeating its liveness record and prevent other nodes
from taking over its leases, despite being completely unresponsive.
—
You are receiving this because you were mentioned.
Reply to this email directly, view it on GitHub
<#5819?email_source=notifications&email_token=ACQSGZ7AXGRGIKBUZ2XGSMTQTF5FTA5CNFSM4JLMNMWKYY3PNVWWK3TUL52HS4DFVREXG43VMVBW63LNMVXHJKTDN5WW2ZLOORPWSZGOEDXHLPY#issuecomment-552498623>,
or unsubscribe
<https://github.com/notifications/unsubscribe-auth/ACQSGZ4EXRRJXSGT26KTAWLQTF5FTANCNFSM4JLMNMWA>
.
--
Alfonso
|
|
Spas Bojanov:
- #40288 <cockroachdb/cockroach#40288> jobs:
make new index backfill status tracking bwd compatible. Release note (bug
fix): this was introduced by #39773 which will cause backfill to crash when
running on cluster with upgraded nodes.
- #39773 <cockroachdb/cockroach#39773> jobs:
improve the status reporting of index backfill jobs. Release note (bug
fix): fixes #36601.
- #41172 <cockroachdb/cockroach#41172> jobs:
clean running jobs with no mutations left. Release note (bug fix): fixes
#40563.
- #40345 <cockroachdb/cockroach#40345> importccl:
fix mysqqoutfile import bug. Release note (bug fix): When importing from
csv if the number of fields is less than the columns in the schema the
server will panic.
On Tue, Nov 12, 2019 at 12:22 AM Alfonso Subiotto Marqués <
notifications@github.com> wrote:
… - #39662 <cockroachdb/cockroach#39662> workload:
fall back to using 19.1 vectorize setting on syntax error. Release note
(bug fix): The querybench workload has been modified to use the correct
syntax for vectorized execution in 19.1, attempting to run this workload
with vectorized execution would previously return an error.
- #41467 <cockroachdb/cockroach#41467>
release-19.2:
colrpc: downgrade log messages from Error to Warning. Release note (bug
fix): Some log messages reported as Errors have been downgraded to
Warnings.
- #37088 <cockroachdb/cockroach#37088> exec:
wrap hashRouter work with CatchVectorizedRuntimeError. Release note (bug
fix): An expected panic in the vectorized engine is now reported as an
error without crashing the server.
- #39093 <cockroachdb/cockroach#39093> exec:
wait for internal goroutine completion in UnorderedSynchronizer. Release
note (bug fix): A race condition in the vectorized engine has been fixed.
- #38006 <cockroachdb/cockroach#38006> exec:
set output columnTypes when creating noop operator. Release note (bug fix):
A bug planning noop operators in the vectorzied execution engine has been
fixed.
- #40668 <cockroachdb/cockroach#40668> exec:
reset internal unsafeBatch in orderedAggregator. Release note (bug fix): A
bug with vectorized aggregations due to incorrect state transitions has
been fixed.
- #38557 <cockroachdb/cockroach#38557> exec:
protect against unset syncFlowConsumer. Release note (bug fix): A
vectorized engine plan-time state assertion was added to avoid a panic.
- #39248 <cockroachdb/cockroach#39248> exec:
fix metadata drain cycles. Release note (bug fix): A bug that led to
queries hanging in the vectorized engine due to incorrect plans has been
fixed.
- #37571 <cockroachdb/cockroach#37571> exec:
check for unset output columnTypes. Release note (bug fix): A vectorized
engine plan-time state assertion was added.
- #38329 <cockroachdb/cockroach#38329> distsqlrun:
include StartScan stall time in EXPLAIN ANALYZE. Release note (sql change):
EXPLAIN ANALYZE now reports the time taken to look up rows in a lookup
join.
- #38622 <cockroachdb/cockroach#38622> distsqlrun:
bump index joiner batch size from 100 to 10k. Release note (performance
improvement): Index join performance has been improved.
- #37745 <cockroachdb/cockroach#37745>
col{data,serde}:
various fixes related to TestArrowBatchConverterRandom (2 commits). Release
note (bug fix): A bug in data serialization in the vectorized engine caused
by incorrectly determining the presence of nulls in rows has been fixed.
On Mon, Nov 11, 2019 at 10:48 AM irfan sharif ***@***.***>
wrote:
>
> - #41764: Release note (bug fix): Previously a disk stall could allow
> a node to continue heartbeating its liveness record and prevent other
nodes
> from taking over its leases, despite being completely unresponsive.
>
> —
> You are receiving this because you were mentioned.
> Reply to this email directly, view it on GitHub
> <
#5819?email_source=notifications&email_token=ACQSGZ7AXGRGIKBUZ2XGSMTQTF5FTA5CNFSM4JLMNMWKYY3PNVWWK3TUL52HS4DFVREXG43VMVBW63LNMVXHJKTDN5WW2ZLOORPWSZGOEDXHLPY#issuecomment-552498623
>,
> or unsubscribe
> <
https://github.com/notifications/unsubscribe-auth/ACQSGZ4EXRRJXSGT26KTAWLQTF5FTANCNFSM4JLMNMWA
>
> .
>
--
Alfonso
—
You are receiving this because you were mentioned.
Reply to this email directly, view it on GitHub
<#5819?email_source=notifications&email_token=AANL635GCJRAF2GECKWGYBDQTGBF7A5CNFSM4JLMNMWKYY3PNVWWK3TUL52HS4DFVREXG43VMVBW63LNMVXHJKTDN5WW2ZLOORPWSZGOEDXKWQQ#issuecomment-552512322>,
or unsubscribe
<https://github.com/notifications/unsubscribe-auth/AANL637KSPBTOLESRUX35EDQTGBF7ANCNFSM4JLMNMWA>
.
|
|
|
I generally think release notes should be as comprehensive as possible. They're not necessarily read, but it's important that they be searchable. If the notes are too large (as they are for a full release), this should be addressed with a separate summary document instead of dropping details. (I like python's what's new/changelog split) |
|
|
* #41571 (cli change) `cockroach debug unsafe-remove-dead-replicas` can now
recover ranges with more than one remaining live replica
* #41164 (bug fix) fixed a bug when trying to backup to local drive in the
Windows version
* #37168 (bug fix) fixed a race condition affecting tests where the node
status is not written before the node starts
* #41160 (bug fix) fixed a rocksdb bug where the `file not found` error
message is interpreted incorrectly under Windows
* #41624 (bug fix) fixed an allocator issue where a range upreplicates to
even number of replicas
…On Mon, Nov 11, 2019 at 9:36 AM Ben Darnell ***@***.***> wrote:
Done, but I think we should have a conversation about what changes are
worthy to be included in release notes. For one, I enjoy reading release
notes for software, but not if they're endless. I'm afraid that activism
like this, coupled with more and more monkeys banging on this, will lead to
a million bullets of very varying degrees of importance.
I generally think release notes should be as comprehensive as possible.
They're not necessarily *read*, but it's important that they be
*searchable*. If the notes are too large (as they are for a full
release), this should be addressed with a separate summary document instead
of dropping details. (I like python's what's new
<https://docs.python.org/3/whatsnew/3.8.html>/changelog
<https://docs.python.org/3/whatsnew/changelog.html#changelog> split)
—
You are receiving this because you were mentioned.
Reply to this email directly, view it on GitHub
<#5819?email_source=notifications&email_token=ABVZ3FKE7L2UF2YSKRPO3B3QTGJYLA5CNFSM4JLMNMWKYY3PNVWWK3TUL52HS4DFVREXG43VMVBW63LNMVXHJKTDN5WW2ZLOORPWSZGOEDXRPYA#issuecomment-552540128>,
or unsubscribe
<https://github.com/notifications/unsubscribe-auth/ABVZ3FNXQSZKAJOFI554DHTQTGJYLANCNFSM4JLMNMWA>
.
|
N/A: #39383 (no user-facing changes) FWIW, with the exception of #36032 and #35742, all of these were bug fixes for bugs introduced in earlier alpha/beta releases, and therefore don't represent differences from 19.1. It's not clear to me what the plan is for incorporating these release notes into the 19.2.0 release notes, but I worry that adding release notes that were bug fixes from earlier alphas/betas indiscriminately to all the other release notes will be misleading. |
|
#39154 (cli change): In the CLI, creating a non-partitioned index on a partitioned table returns an error unless sql_safe_updates = false. I omitted PRs which didn't represent a user-visible change from 19.1 or weren't really of interest to users, like adding a HINT to an error message. |
Not adding notes for #40757, since that was a fix for a PR/feature that was added in this release, and that feature already has release notes. Also, I noticed that the |
|
|
|
|
@knz, @bdarnell, should we retroactively add these notes to their relevant release note pages? If so, I'll need authors to identify the releases. I suspect that's too much work, but I'm open to your guidance. More importantly, once all authors have completed their reviews here, I'll open docs issues, where relevant. |
|
If ben and you decide that retroactive addition is the way to go, then we can automate the classification. That is not too much work. However I am not sure this is the way to go? Or perhaps, not the only place where this should be reported: all the users that have already read the past release notes will not come across this new content. Maybe we should instead publish an "erratum" page specific to 19.2, even communicate it on the forum and perhaps other channels, and then link to that page from the other release note pages for reference. |
|
I meant "addendum" not "erratum". |
|
Whatever we do here should be fairly quiet. https://www.cockroachlabs.com/docs/releases/v19.2.0.html is already a summary of the release, and if we were going to have a more detailed "addendum" it would primarily include the union of all the alpha/beta/rc release notes before it would include these things barely got mentioned. It doesn't make sense to make a new page for these. Fitting these into their corresponding alpha/etc release notes is the only place we have to put these notes today and I'm not sure it's worth the effort. Going forward, perhaps we should consider going to a model where minor releases are lumped together on one page, so we'd have three pages per branch: one for all the pre-releases of a branch (which would also serve as the detailed changelog for the release and would be a reasonable place to drop in unsorted "forgotten" changes like these), one for the release summary itself, and one for the post-release patches. |
jseldess
added
A-general
O-eng
P-1
|
It would be nice to have a link to this issue/s from somewhere in docs. I faced: Looked into release notes and didn't found anything related. Took me a while to land here. |
|
I'm closing this issue in favor of discrete issues, which I will create soon. |
and others
Pursuant to cockroachdb/cockroach#42345 I have scanned the entire 19.2 commit history (automatically) to pick up PRs where the author blindly used 'Release note: None' while there were user-facing changes. These are the changes that likely did not show up in release notes, and the doc team likely could not pick up the change to be documented.
If some of your changes are listed here, please review the PR(s) and suggest missing release notes + doc changes as a comment below.
Also don't forget the category!
Then click the checkbox on the left to indicate you've handled the change.
Note: I did a manual filtering on the automatic scan to remove false positives (no release note actually needed) but I may have left some. Mark the checkbox after you've confirmed that the PR indeed contains no user-facing change.
Changes without release note annotation
PRs merged by contributors
Hall of fame: all user-facing changes are properly documented
Honorable mentions: just a few release note omissions
Aayush Shah: @aayushshah15
Andy Kimball: @andy-kimball
Arul Ajmani: @arulajmani
Bilal Akhtar: @itsbilal
Bram Gruneir: @BramGruneir
George Papadrosou: @tbg
Irfan Sharif: @irfansharif
Pete Vilter: @vilterp
Peter Mattis: @petermattis
Ridwan Sharif: @jordanlewis
UPDATE ... FROMstatements (3 commits)Roland Crosby: @rolandcrosby
psqlTaufiq Rahman: @andreimatei
Vivek Menezes: @lucy-zhang
Missing release notes (rest)
Aditya Maru: @jordanlewis
Alfonso Subiotto Marqués: @asubiotto
Andrei Matei: @andreimatei
Andrew Kryczka: @ajkr
Andrew Werner: @ajwerner
Arseni Lapunov: @jordanlewis
Daniel Harrison: @danhhz
fixtures importtooBen Darnell, Darin Peshev: @bdarnell @darinpp
Darin Peshev: @darinpp
David Taylor: @dt
Jeffrey Xiao: @jordanlewis
cockroach demoJordan Lewis, Lucy Zhang: @jordanlewis @lucy-zhang
Jordan Lewis, Nathan VanBenschoten: @jordanlewis @nvanbenschoten
Jordan Lewis: @jordanlewis
Justin Jaffray: @justinj
Lucy Zhang: @lucy-zhang
Nathan VanBenschoten: @nvanbenschoten
Nikhil Benesch, Raphael 'kena' Poss: @knz
Nikhil Benesch: @knz @bobvawter
Paul Bardea: @pbardea
Radu Berinde: @RaduBerinde
Rafi Shamim: @rafiss
Raphael 'kena' Poss: @knz
Rebecca Taft: @rytaft
Rohan Yadav: @rohany
Sean Loiselle: @nvanbenschoten
Solon Gordon: @solongordon
Spas Bojanov: @spaskob
Tobias Schottdorf: @tbg
Yahor Yuzefovich: @yuzefovich
The text was updated successfully, but these errors were encountered: