SnapshotTransaction stream refactor proof-of-concept #7003

wi11dey · 2024-03-04T17:15:00Z

Could probably also separate out all the postfiltering logic into its own PostFilter class

Birthday present for @rhuffy

General

Before this PR:

After this PR:

==COMMIT_MSG==
==COMMIT_MSG==

Priority:

Concerns / possible downsides (what feedback would you like?):

Is documentation needed?:

Compatibility

Does this PR create any API breaks (e.g. at the Java or HTTP layers) - if so, do we have compatibility?:

Does this PR change the persisted format of any data - if so, do we have forward and backward compatibility?:

The code in this PR may be part of a blue-green deploy. Can upgrades from previous versions safely coexist? (Consider restarts of blue or green nodes.):

Does this PR rely on statements being true about other products at a deployment - if so, do we have correct product dependencies on these products (or other ways of verifying that these statements are true)?:

Does this PR need a schema migration?

Testing and Correctness

What, if any, assumptions are made about the current state of the world? If they change over time, how will we find out?:

What was existing testing like? What have you done to improve it?:

If this PR contains complex concurrent or asynchronous code, is it correct? The onus is on the PR writer to demonstrate this.:

If this PR involves acquiring locks or other shared resources, how do we ensure that these are always released?:

Execution

How would I tell this PR works in production? (Metrics, logs, etc.):

Has the safety of all log arguments been decided correctly?:

Will this change significantly affect our spending on metrics or logs?:

How would I tell that this PR does not work in production? (monitors, etc.):

If this PR does not work as expected, how do I fix that state? Would rollback be straightforward?:

If the above plan is more complex than “recall and rollback”, please tag the support PoC here (if it is the end of the week, tag both the current and next PoC):

Scale

Would this PR be expected to pose a risk at scale? Think of the shopping product at our largest stack.:

Would this PR be expected to perform a large number of database calls, and/or expensive database calls (e.g., row range scans, concurrent CAS)?:

Would this PR ever, with time and scale, become the wrong thing to do - and if so, how would we know that we need to do something differently?:

Development Process

Where should we start reviewing?:

If this PR is in excess of 500 lines excluding versions lock-files, why does it not make sense to split it?:

Please tag any other people who should be aware of this PR:

changelog-app · 2024-03-04T17:15:04Z

Generate changelog in `changelog/@unreleased`

What do the change types mean?

feature: A new feature of the service.
improvement: An incremental improvement in the functionality or operation of the service.
fix: Remedies the incorrect behaviour of a component of the service in a backwards-compatible way.
break: Has the potential to break consumers of this service's API, inclusive of both Palantir services
and external consumers of the service's API (e.g. customer-written software or integrations).
deprecation: Advertises the intention to remove service functionality without any change to the
operation of the service itself.
manualTask: Requires the possibility of manual intervention (running a script, eyeballing configuration,
performing database surgery, ...) at the time of upgrade for it to succeed.
migration: A fully automatic upgrade migration task with no engineer input required.

Note: only one type should be chosen.

How are new versions calculated?

❗The break and manual task changelog types will result in a major release!
🐛 The fix changelog type will result in a minor release in most cases, and a patch release version for patch branches. This behaviour is configurable in autorelease.
✨ All others will result in a minor version release.

Type

Description

SnapshotTransaction stream refactor proof-of-concept

Check the box to generate changelog(s)

Generate changelog entry

rhuffy · 2024-03-04T17:31:49Z

...sdb-impl-shared/src/main/java/com/palantir/atlasdb/transaction/impl/SnapshotTransaction.java

+                    .thenApply(remainingRawResults -> getWithPostFilteringInternal(
+                            tableRef,
+                            remainingRawResults,
+                            asyncKeyValueService,
+                            asyncTransactionService,
+                            iterations + 1))


recursion ⚠️

there's a max iterations precondition already so no danger of stack overflow. the values on the stack are just references to the result maps that need to be kept alive during collecting postfiltered results in any case

rhuffy · 2024-03-04T17:32:24Z

...sdb-impl-shared/src/main/java/com/palantir/atlasdb/transaction/impl/SnapshotTransaction.java

+        Map<Cell, Value> filteredResults = EntryStream.of(rawResults)
+                .removeKeys(keysToDelete::containsKey)
+                .removeKeys(keysToReload::containsKey)
+                .removeValues(value -> value.getContents().length == 0)
+                // The value has a commit timestamp less than our start timestamp, and is visible and valid.
+                .collect(MoreCollectors.entriesToCustomMap(LinkedHashMap::new));


Extra allocation should be avoided

rhuffy · 2024-03-04T17:32:41Z

...sdb-impl-shared/src/main/java/com/palantir/atlasdb/transaction/impl/SnapshotTransaction.java

+        Map<Cell, Long> keysToDelete = EntryStream.of(rawResults)
+                .mapValues(Value::getTimestamp)
+                .filterKeyValue((key, ts) -> commitTimestamps.getIfAbsent(ts, TransactionConstants.FAILED_COMMIT_TS)
+                        == TransactionConstants.FAILED_COMMIT_TS)
+                .toImmutableMap();
+        Map<Cell, Long> keysToReload = EntryStream.of(rawResults)
+                .removeKeys(keysToDelete::containsKey)
+                .mapValues(Value::getTimestamp)


double iteration should be avoided

rhuffy · 2024-03-04T17:34:27Z

As an exercise I think this is fine, but to actually merge this, we'd probably need more than just "streams are easier to read" as a justification, especially if there's additional allocation cost in the hot read path.

jeremyk-91 · 2024-03-04T17:46:02Z

Could probably also separate out all the postfiltering logic into its own PostFilter class

Yep, there's actually an internal project ongoing to slice this into many parts (see #7000 or internal RFCs). I'm curious if there was something that motivated this though?

Birthday present for @rhuffy

🎉

rhuffy · 2024-03-04T18:03:26Z

I was talking to Will about this refactor in general, and how some of the control flow (specifically the use of a resultsAccumulator) is a bit confusing for the uninitiated.

wi11dey · 2024-03-04T18:50:59Z

yup, motivation was just a proof-of-concept since raymond and i were discussing the code in the context of db upgrading and that it was difficult to plug in the kvs migrator, and also the code was a bit hard to follow for the uninitiated because of the various future chaining. i suggested streams and either raymond or jakub said it wouldn't be possible to use them async so this was a proof of concept using streams while keeping async behavior

the extra allocation of keysToDeleteOrReload could be removed with a map view -- we were already allocating for the result map, just renamed it to filteredResults. the extra iterations could probably also be removed with a little effort

I think this version is easier to reason about since there's no bouncing back and forth between getWithPostFilteringIterate/getWithPostFilteringInternal, i believe @Output in general should be avoided when possible so that the flow of values matches the code flow, and you need to keep less state in your head while reading through the code

anyway not suggesting we merge this right now, but the Future Spliterator approach could be something to keep in mind/reference if this code is going to be refactored anyway

wi11dey added 10 commits March 2, 2024 22:47

@output is unidiomatic and unnecessary in SnapshotTransaction

eff4f94

@DoNotDelegate

743b02a

cleanup

09439e7

concision

7844158

fix test

f2135a1

deps

221eb11

nonNullValues

e5e454a

more cleanup

8eb187e

update delete & rollback conditional

f32e045

unnecessary putAll

bc561e7

rhuffy reviewed Mar 4, 2024

View reviewed changes

rhuffy mentioned this pull request Mar 4, 2024

[WIP] KeyValueSnapshotReaderFactory #7004

Closed

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

SnapshotTransaction stream refactor proof-of-concept #7003

SnapshotTransaction stream refactor proof-of-concept #7003

wi11dey commented Mar 4, 2024 •

edited

Loading

changelog-app bot commented Mar 4, 2024

rhuffy Mar 4, 2024

wi11dey Mar 4, 2024 •

edited

Loading

rhuffy Mar 4, 2024

rhuffy Mar 4, 2024

rhuffy commented Mar 4, 2024

jeremyk-91 commented Mar 4, 2024

rhuffy commented Mar 4, 2024

wi11dey commented Mar 4, 2024 •

edited

Loading

SnapshotTransaction stream refactor proof-of-concept #7003

Are you sure you want to change the base?

SnapshotTransaction stream refactor proof-of-concept #7003

Conversation

wi11dey commented Mar 4, 2024 • edited Loading

General

Compatibility

Testing and Correctness

Execution

Scale

Development Process

changelog-app bot commented Mar 4, 2024

Generate changelog in changelog/@unreleased

rhuffy Mar 4, 2024

Choose a reason for hiding this comment

wi11dey Mar 4, 2024 • edited Loading

Choose a reason for hiding this comment

rhuffy Mar 4, 2024

Choose a reason for hiding this comment

rhuffy Mar 4, 2024

Choose a reason for hiding this comment

rhuffy commented Mar 4, 2024

jeremyk-91 commented Mar 4, 2024

rhuffy commented Mar 4, 2024

wi11dey commented Mar 4, 2024 • edited Loading

wi11dey commented Mar 4, 2024 •

edited

Loading

Generate changelog in `changelog/@unreleased`

wi11dey Mar 4, 2024 •

edited

Loading

wi11dey commented Mar 4, 2024 •

edited

Loading