Recognize byte slice for key argument in cluster client hash slot computation #3049
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
|
It seems that you have restricted modifications to your PR code. You can either open the permissions or resolve the conflicts yourself. |
e1f8ca8 to
7e90536
Compare
Done! Rebased and conflicts resolved |
monkey92t
approved these changes
Jul 16, 2024
|
@ofekshenawa merge? |
|
Is it possible to merge this? This PR fixes a pretty significant performance bug for us. Thanks! |
ofekshenawa
approved these changes
Jul 23, 2024
35ae45e to
3cd7c33
Compare
|
@ofekshenawa I rebased this branch and the build is now green. Are we good to merge? Thanks. |
|
@vladvildanov @ofekshenawa Are you able to help with this merge? We have been running this patch in production for several months and have confirmed that it fixes the issue in the PR description. |
|
@LINKIWI Sorry for the delay, approved and merged! |
Merged
ndyakov
pushed a commit
that referenced
this pull request
Feb 21, 2025
…putation (#3049) Co-authored-by: Vladyslav Vildanov <117659936+vladvildanov@users.noreply.github.com> Co-authored-by: ofekshenawa <104765379+ofekshenawa@users.noreply.github.com>
ndyakov
added a commit
that referenced
this pull request
Feb 21, 2025
* Add guidance on unstable RESP3 support for RediSearch commands to README (#3177) * Add UnstableResp3 to docs * Add RawVal and RawResult to wordlist * Explain more about SetVal * Add UnstableResp to wordlist * Eliminate redundant dial mutex causing unbounded connection queue contention (#3088) * Eliminate redundant dial mutex causing unbounded connection queue contention * Dialer connection timeouts unit test --------- Co-authored-by: ofekshenawa <104765379+ofekshenawa@users.noreply.github.com> * SortByWithCount FTSearchOptions fix (#3201) * SortByWithCount FTSearchOptions fix * FTSearch test fix * Another FTSearch test fix * Another FTSearch test fix --------- Co-authored-by: Christopher Golling <Chris.Golling@aexp.com> * Fix race condition in clusterNodes.Addrs() (#3219) Resolve a race condition in the clusterNodes.Addrs() method. Previously, the method returned a reference to a string slice, creating the potential for concurrent reads by the caller while the slice was being modified by the garbage collection process. Co-authored-by: Nedyalko Dyakov <nedyalko.dyakov@gmail.com> * chore: fix some comments (#3226) Signed-off-by: zhuhaicity <zhuhai@52it.net> Co-authored-by: Nedyalko Dyakov <nedyalko.dyakov@gmail.com> * fix(aggregate, search): ft.aggregate bugfixes (#3263) * fix: rearange args for ft.aggregate apply should be before any groupby or sortby * improve test * wip: add scorer and addscores * enable all tests * fix ftsearch with count test * make linter happy * Addscores is available in later redisearch releases. For safety state it is available in redis ce 8 * load an apply seem to break scorer and addscores * fix: add unstableresp3 to cluster client (#3266) * fix: add unstableresp3 to cluster client * propagate unstableresp3 * proper test that will ignore error, but fail if client panics * add separate test for clusterclient constructor * fix: flaky ClientKillByFilter test (#3268) * Reinstate read-only lock on hooks access in dialHook (#3225) * use limit when limitoffset is zero (#3275) * remove redis 8 comments * update package versions * use latest golangci-lint * fix(search&aggregate):fix error overwrite and typo #3220 (#3224) * fix (#3220) * LOAD has NO AS param(https://redis.io/docs/latest/commands/ft.aggregate/) * fix typo: WITHCOUT -> WITHCOUNT * fix (#3220): * Compatible with known RediSearch issue in test * fix (#3220) * fixed the calculation bug of the count of load params * test should not include special condition * return errors when they occur --------- Co-authored-by: Nedyalko Dyakov <nedyalko.dyakov@gmail.com> Co-authored-by: ofekshenawa <104765379+ofekshenawa@users.noreply.github.com> * Recognize byte slice for key argument in cluster client hash slot computation (#3049) Co-authored-by: Vladyslav Vildanov <117659936+vladvildanov@users.noreply.github.com> Co-authored-by: ofekshenawa <104765379+ofekshenawa@users.noreply.github.com> --------- Signed-off-by: zhuhaicity <zhuhai@52it.net> Co-authored-by: ofekshenawa <104765379+ofekshenawa@users.noreply.github.com> Co-authored-by: LINKIWI <LINKIWI@users.noreply.github.com> Co-authored-by: Cgol9 <chris.golling@verizon.net> Co-authored-by: Christopher Golling <Chris.Golling@aexp.com> Co-authored-by: Shawn Wang <62313353+shawnwgit@users.noreply.github.com> Co-authored-by: ZhuHaiCheng <zhuhai@52it.net> Co-authored-by: herodot <54836727+bitsark@users.noreply.github.com> Co-authored-by: Vladyslav Vildanov <117659936+vladvildanov@users.noreply.github.com>
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
This PR is mildly related to #3048.
The cluster client uses the command type to determine the expected key position in the command arguments slice to compute the hash slot. This enables the client to properly send the command to the node that owns that slot, according to the cached cluster topology state.
For generic commands assembled with the
Do()API, the logic falls back to assuming the key is at position 1. Though this is a reasonable assumption, it depends on the argument to be ofstringtype;[]bytetype arguments are not converted to strings, which results in an incorrect slot computation, thus forcing (in most cases) a double round trip caused by response to aMOVEDredirection.This PR adds a
[]bytetype case for key string serialization, so that the generic commands passed as byte slices yield the correct cluster hash slot.