Catchup to upstream - #2
Open
Bernie wants to merge 60 commits into
Open
Conversation
…ssing an out-of-range slot (redis-rb#466)
…hRing Ref: rails/rails#57004 Ref: redis-rb/redis-client#292 I'm working on a big cleanup of Rails `RedisCacheStore` and it needs to support consistent hashing, which I did add in `redis-client` `0.28.0`. For Rails users to be able to use a redis cluster as a cache store, `redis-cluster-client` would need to have a shim for the now base interface: - `with` - `nodes` - `node_for` - `nodes_for`
Co-authored-by: Copilot Autofix powered by AI <62310815+github-advanced-security[bot]@users.noreply.github.com>
…uster-state change (redis-rb#476)
Fix typos (metdata, hashag, guaranted, transction, fhe, GItHub), grammar issues (missing verbs, wrong verb forms, uncountable nouns), and Japanese-English patterns (upper libraries, works multiple keys, recommend to use, the followings, etc.) across README.md and source comments. Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
…mplementation
- class_diagrams_redis_cluster_client.md:
- RedisClient::Cluster: add multi() and with()
- RedisClient::ClusterConfig: replace obsolete per_node_key/update_node/add_node/dup
with client_config_for_node, resolved?, sentinel?, server_url, and timeout readers
- RedisClient::Cluster::Command: add exists?()
- RedisClient::Cluster::Node: remove non-existent self.load_info(), add try_reload!,
clients, primary_clients, replica_clients
- Add RedisClient::Cluster::Node::BaseTopology class
- Add RedisClient::Cluster::Node::RandomReplicaOrPrimary class (was missing)
- Remove module_RedisClient_Cluster_Node_ReplicaMixin (no longer exists)
- Update topology class relationships to use inheritance from BaseTopology
- RedisClient::Cluster::Router: remove non-existent try_send(), add all current
public methods (handle_redirection, scan_single_key, find_node_key_by_key, etc.)
- Add RedisClient::Cluster::OptimisticLocking class and its relationships
- module_RedisClient_Cluster_NodeKey: add build_from_client()
- class_diagrams_redis_client.md:
- RedisClient: add self.ring, self.now, self.now_ms, server_url, db, host, port,
path, username, password, idle_timeout, disable_reconnection, measure_round_trip_delay
- module_RedisClient_Common: add nodes, node_for, nodes_for
- module_RedisClient_Config_Common: add idle_timeout, protocol, circuit_breaker,
custom, inherit_socket, driver_info, middlewares_stack, resolved?, server_url,
build_lib_name
- RedisClient::SentinelConfig: add resolved?
- RedisClient::RubyConnection: add measure_round_trip_delay
- sequence_diagrams.md:
- Update initial topology fetch from CLUSTER NODES to CLUSTER SHARDS
(with fallback note for older Redis versions)
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Co-authored-by: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
ClusterConfig#inspect previously dumped startup_nodes hashes which contained raw :username and :password values, leaking them into logs, exception messages, IRB output, APM traces, etc. Replace sensitive values with '[FILTERED]' before constructing the inspect string. Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Co-authored-by: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
…nError The retry counter in handle_redirection used [times_block_executed, 1].min, which deducts 0 when the inner block never executed (e.g., when initial node lookup or pre-block code in Router#handle_redirection raises). The inline comment said "Always deduct at least one"; the code did the opposite. Persistent ConnectionError on the WATCH path could spin without ever decrementing retry_count. Switching to .max matches the documented intent. Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
NodeKey.split previously used rindex(':') to separate host from
port. This works for bracket-less IPv6 ('::1:6379') but leaves
the brackets attached for the standard URL form ('[::1]:6379'),
producing a host '[::1]' that downstream connection code cannot
resolve.
Detect the [host]:port form explicitly and strip the brackets.
The legacy bracket-less path is preserved unchanged.
build_from_host_port is intentionally left as-is in this PR;
changing the produced NodeKey format has cluster-wide implications
and belongs in a separate change.
PubSub#@commands logged every command and was replayed in start_over on reconnect. Subscribe entries for channels the user later unsubscribed from were never removed, so long-lived clients accumulated unbounded memory and replays were O(history) instead of O(active subscriptions). Track only the active subscription set: subscribe-family commands are appended; unsubscribe-family commands prune matching entries (by command type and channel list) and are not themselves stored.
Pipeline#execute previously raised v.first_exception eagerly on the first node-batch that carried one, dropping redirection processing and cluster_state_errors processing for all remaining batches. Defer the raise until after all batches have been processed; preserve the original "raise the first encountered exception" semantics so the exception type seen by the caller is unchanged.
Some managed Redis services (e.g. AWS ElastiCache Serverless) report
`127.0.0.1` in the `ip` field of `CLUSTER SHARDS` while exposing the
real, reachable address only via `endpoint` / `hostname`. Building
node_keys from `ip` then yields an unreachable topology and the
client fails on the next operation that walks all nodes (e.g.
`SCAN`/`delete_matched`) with:
Connection refused - connect(2) for 127.0.0.1:6379
This regressed in v0.16.0 (redis-rb#479) when topology discovery switched
from `CLUSTER NODES` to `CLUSTER SHARDS`; redis-rb#207 had previously
established hostname precedence for the `CLUSTER NODES` path.
`parse_cluster_shards_reply` now picks the host using:
1. `endpoint` (server-selected per `cluster-preferred-endpoint-type`)
2. `hostname` (when `endpoint` is absent or `?`)
3. `ip` (fallback)
mirroring the precedence used by `parse_node_key` for CLUSTER NODES.
Refs: redis-rb#515
Bernie
force-pushed
the
catchup_to_upstream
branch
from
August 13, 2026 14:53
61613a0 to
cd950a0
Compare
This file contains hidden or 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
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
4 participants
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 just catches our master up to the upstream.