Releases: nats-io/nats-streaming-server
Releases · nats-io/nats-streaming-server
Release v0.24.1
Changelog
Note that we added a deprecation notice for this project. See note here
Go Version
- 1.17.6: Both release executables and Docker images are built with this Go release.
Updated
- Dependencies (#1238)
- NATS Server v2.7.1 -> v2.7.2
Fixed
- Possible panic when monitoring endpoints are inspected while the server is in the middle of recovering its state. Thanks to @wilstoff for the report (#1236)
- Possible panic on message redelivery (#1237)
Complete Changes
Release v0.24.0
Changelog
Note that we added a deprecation notice for this project. See note here
Go Version
- 1.17.6: Both release executables and Docker images are built with this Go release.
Updated
Changed
- The
/vendor
directory has been removed and the executable is now built with go modules (#1233) - The error messages "request missing subject", "request for missing inbox" and "request for missing client" have been respectively replaced by "request for unknown channel", "request for unknown inbox" and "request for unknown client". This is mentioned here in case some users were filtering on those error messages (#1231)
Fixed
- Clustering:
- Link to documentation in README.md. Thanks to @EJTJ3 for the contribution (#1227)
- Missing log statement when failing to create a channel (#1232)
Complete Changes
Release v0.23.2
Changelog
Note that we added a deprecation notice for this project. See note here
Go Version
- 1.16.10: Both release executables and Docker images are built with this Go release.
Updated
- Dependencies (#)
- NATS Server v2.6.4 -> v2.6.5
Fixed
- Clustering:
- Possible wrong
pending_count
andis_stalled
in followers. This was introduced in v0.23.0. Thanks to @bfoxstudio for the report (#1220)
- Possible wrong
Complete Changes
Release v0.23.1
Changelog
Note that we added a deprecation notice for this project. See note here
Go Version
- 1.16.10: Both release executables and Docker images are built with this Go release.
Updated
- Dependencies (#)
- Raft v1.3.1 -> v1.3.2
- Postgres v1.10.3 -> v1.10.4
- NATS Server v2.6.2 -> v2.6.4
Fixed
- Clustering:
- Possible wrong pending count. This was introduced in v0.23.0. Thanks to @mihai-tiriplica-f3 for the report (#1220)
Complete Changes
Release v0.23.0
Changelog
Note that we added a deprecation notice for this project. See note here
Go Version
- 1.16.9: Both release executables and Docker images are built with this Go release.
Updated
- Dependencies (#1215)
- Postgres v1.10.2 -> v1.10.3
- NATS Server v2.3.3 -> v2.6.2
Improved
- Clustering:
- Redelivery for queue members is now similar to standalone/FT mode, which means that messages will be redelivered to any member of the group (as opposed to the member that originally received the messages). Thanks to @mihai-tiriplica-f3 for the suggestion (#1214)
Fixed
- Clustering:
Complete Changes
Release v0.22.1
Changelog
Note that we added a deprecation notice for this project. See note here
Also, the master
branch was renamed to main
. If you have a fork/clone of the repo, you may have to run this:
git branch -m master main
git fetch origin
git branch -u origin/main main
git remote set-head origin -a
Go Version
- 1.16.6: Both release executables and Docker images are built with this Go release.
Updated
- Dependencies (#1206)
- NATS Server v2.2.6 -> v2.3.3
- bbolt v1.3.5 -> v1.3.6
Changed
- Executable symbol table no longer stripped (#1206)
Fixed
- Clustering:
- Possible redelivery from a node that acquired leadership and lost it quickly after. Thanks to @piotrlipiarz-ef for the report (#1201)
- Snapshot of queue members would not capture the
last_sent
value of the group, which could possibly lead to redelivery of old messages when a node became leader. Thanks to @mihai-tiriplica-f3 for the report (#1203)
- SQL:
- A subscription's
last_sent
could end-up higher than the store'slast_seq
in presence of message expiration and server restarts (standalone or Fault Tolerance modes), which would cause the server to stop sending messages until the message sequences were above the subscription'slast_sent
(#1204)
- A subscription's
Complete Changes
Release v0.22.0
Changelog
Note that we added a deprecation notice for this project. See note here
Go Version
- 1.16.4: Both release executables and Docker images are built with this Go release.
Added
Updated
- Dependencies (#1192, #1193, #1196)
- go-sql-driver/mysql v1.5.0 to v1.6.0
- hashicorp/raft v1.2.0 to v1.3.1
- lib/pq v1.9.0 to v1.10.2
- nats-io/nats-serer v2.1.9 to v2.2.6
- nats-io/nats.go v1.10.0 to v1.11.0
Fixed
- Clustering
- Possible panic on shutdown (#1184)
- Subscription redelivery count map was not cleaned-up when processing ACK from subscription, which may lead to memory growth if there were a lot of redelivered messages (#1190)
Complete Changes
Release v0.21.2
Changelog
Note that we added a deprecation notice for this project. See note here
Go Version
- 1.16.3: Both release executables and Docker images are built with this Go release.
Fixed
- Clustering: unexpected memory growth. Thanks to @hullarb, @kkoppel and @sergeiepam for the report (#1177)
Complete Changes
Release v0.21.1
Changelog
Go Version
- 1.16: Both release executables and Docker images are built with this Go release.
Fixed
- Clustering: Unlikely panic with RAFT transport, introduced in v0.20.0 (#1167)
- Clustering: Possible slow consumer errors with RAFT transport, introduced in v0.20.0 (#1169)
- Clustering: Added missing node_id in
/streaming/serverz
endpoint. Thanks to @sskserk and @sergeiepam for the report (#1171)
Complete Changes
Release v0.21.0
Changelog
Go Version
- 1.16: Both release executables and Docker images are built with this Go release.
Added
- Subscriptions count in some monitoring endpoints (#1154)
- Clustering: configuration parameters allowing configuration of some of bolt DB options. This is use for RAFT storage. The new option
bolt_free_list_map
may help with fragmented free list and big RAFT logs (#1159) - Clustering: configuration parameter
nodes_connections
to enable creation of separate NATS connections to communicate between RAFT nodes (#1160)
Improved
- Clustering: Reduce use of locking where not needed (#1158)
- Reduce contention when sending heartbeats to clients (#1157)
Updated
- Protobuf dependencies (#1146)
Fixed
- Possibly dropping clients protocol requests (#1153)
- Fail pending subscriptions requests if the client connection is closed due to lack of heartbeats (#1161)
- Support closing of subscriptions based on subscription's Inbox instead of AckInbox. New libraries should send a subscription close request if they get a timeout on subscribe (#1162)