mirror of https://github.com/apache/cassandra
ninja-fix CHANGES.txt and NEWS.txt (move entries to 4.1.7)
This commit is contained in:
parent
5e37fcde46
commit
8e0fbc3f28
18
CHANGES.txt
18
CHANGES.txt
|
@ -1,10 +1,4 @@
|
|||
4.1.6
|
||||
* Add native transport deadline, an ultimate deadline for all tasks related to a specific request (CASSANDRA-19534)
|
||||
* Do not create a role if ALTER ROLE IF EXISTS operates on non-existing role (CASSANDRA-19749)
|
||||
* Use OpOrder in repairIterator to ensure we don't lose memtables mid-paxos repair (Cassandra-19668)
|
||||
* Refresh stale paxos commit (CASSANDRA-19617)
|
||||
* Reduce info logging from automatic paxos repair (CASSANDRA-19445)
|
||||
* Support legacy plain_text_auth section in credentials file removed unintentionally (CASSANDRA-19498)
|
||||
4.1.7
|
||||
Merged from 4.0:
|
||||
* Safer handling of out-of-range tokens (CASSANDRA-13704)
|
||||
* Fix memory leak in BTree.FastBuilder (CASSANDRA-19785)
|
||||
|
@ -13,6 +7,16 @@ Merged from 4.0:
|
|||
* Fix rendering UNSET collection types in query tracing (CASSANDRA-19880)
|
||||
* Fix latency reported by ideal consistency level monitoring (CASSANDRA-19651)
|
||||
* Use default commitlog settings in test YAMLs (CASSANDRA-19830)
|
||||
|
||||
|
||||
4.1.6
|
||||
* Add native transport deadline, an ultimate deadline for all tasks related to a specific request (CASSANDRA-19534)
|
||||
* Do not create a role if ALTER ROLE IF EXISTS operates on non-existing role (CASSANDRA-19749)
|
||||
* Use OpOrder in repairIterator to ensure we don't lose memtables mid-paxos repair (Cassandra-19668)
|
||||
* Refresh stale paxos commit (CASSANDRA-19617)
|
||||
* Reduce info logging from automatic paxos repair (CASSANDRA-19445)
|
||||
* Support legacy plain_text_auth section in credentials file removed unintentionally (CASSANDRA-19498)
|
||||
Merged from 4.0:
|
||||
* Do not spam log with SSLExceptions (CASSANDRA-18839)
|
||||
* Fix schema.cql created by a snapshot after dropping more than one column (CASSANDRA-19747)
|
||||
* UnsupportedOperationException when reducing scope for LCS compactions (CASSANDRA-19704)
|
||||
|
|
30
NEWS.txt
30
NEWS.txt
|
@ -51,6 +51,24 @@ restore snapshots created with the previous major version using the
|
|||
'sstableloader' tool. You can upgrade the file format of your snapshots
|
||||
using the provided 'sstableupgrade' tool.
|
||||
|
||||
4.1.7
|
||||
=====
|
||||
|
||||
Upgrading
|
||||
---------
|
||||
- This release introduces safeguards and observability into possible data loss scenarios when nodes have a
|
||||
divergent view of the cluster. This happens around edge-cases on unsafe bootstrapping, decommissions, or when a
|
||||
node has a corrupted topology. Two configuration options have been added: `log_out_of_token_range_requests` and
|
||||
`reject_out_of_token_range_requests`, both enabled by default. The former will make nodes log requests they
|
||||
receive that don't belong in their current or pending token ranges. The latter will reject those requests
|
||||
altogether, which prevents the vast majority of eventual data loss that can occur, but may also incur small
|
||||
windows of degraded availability during range movements in an unhealthy cluster. Both options apply to streaming,
|
||||
mutations, hints, single-partition reads, read-repair, and Paxos operations. Logging these requests is recommended
|
||||
in all situations, and operators are encouraged to be aware when such events occur and to investigate if writes
|
||||
have only been received on incorrect nodes. Operators that are confident of topology parity across all nodes may
|
||||
disable rejection to maximize availability. See CASSANDRA-13704 for further details. Transactional Cluster
|
||||
Metadata in version 5.1 will remove all such edge-cases, providing appropriate guarantees.
|
||||
|
||||
4.1.6
|
||||
=====
|
||||
|
||||
|
@ -60,18 +78,6 @@ Upgrading
|
|||
improved management of pending requests; any custom implementations
|
||||
(using the cassandra.custom_query_handler_class system property) will
|
||||
need to be recompiled. For further details, see CASSANDRA-19534.
|
||||
- This release introduces safeguards and observability into possible data loss scenarios when nodes have a
|
||||
divergent view of the cluster. This happens around edge-cases on unsafe bootstrapping, decommissions, or when a
|
||||
node has a corrupted topology. Two configuration options have been added: `log_out_of_token_range_requests` and
|
||||
`reject_out_of_token_range_requests`, both enabled by default. The former will make nodes log requests they
|
||||
receive that don't belong in their current or pending token ranges. The latter will reject those requests
|
||||
altogether, which prevents the vast majority of eventual data loss that can occur, but may also incur small
|
||||
windows of degraded availability during range movements in an unhealthy cluster. Both options apply to streaming,
|
||||
mutations, hints, single-partition reads, read-repair, and Paxos operations. Logging these requests is recommended
|
||||
in all situations, and operators are encouraged to be aware when such events occur and to investigate if writes
|
||||
have only been received on incorrect nodes. Operators that are confident of topology parity across all nodes may
|
||||
disable rejection to maximize availability. See CASSANDRA-13704 for further details. Transactional Cluster
|
||||
Metadata in version 5.1 will remove all such edge-cases, providing appropriate guarantees.
|
||||
|
||||
4.1.3
|
||||
=====
|
||||
|
|
Loading…
Reference in New Issue