You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
This commit was created on GitHub.com and signed with GitHub’s verified signature.
The key has expired.
release 21.1.4
Major release:
This work includes 420 commits.
Major features:
* Initial feature of coprocessor with coproc wasm (WASM) feature
exposed under `rpk wasm {deploy,remove}` commands.
We note that this feature while in the tree is marked experimental.
Documentation to follow.
* Initial feature of archival (s3) backups.
We landed the metadata description and s3 archival functions
alongside all of the secure hashing schemes needed by s3.
* Cross segmen compaction finally landed in tree which allows for
compaction of adjacent segments. with an initially exposed max file handle size of 5G
for adjacent segment compaction
* RPK landed a huge number of bugfixes and simplifications, most notably namespacing
the new namespacing is
rpk {redpanda, debug, topic} - which should hopefully allow users to
understand RPK better.
* First patch of idempotent producer also landed with full compatibility
targetted for the next release :) - We note that this is the first step in turn
to support transactions.
* A new `kafka::client` written in seastar was ported out of the http proxy
into the top level kafka namespace
* First pass for live partition migration landed, which included node id
virtualization. Need to hook this up to the `controller backend` to
expose it to users.
* This build includes a HUGE amount of performance improvements where
we can now gurantee hardware saturation in all modes of acknowledgement.
* This change also comes with priority-based raft leader election, among
RPC reordering optimization, amortized flushes by a 'look-ahead' flush
on the followers. The TCP clients were increased to 8 from 3 to allow
for saturation on large machines.
* Added GCP specific tuner to `rpk tuner` to maximize throughput on
GCP after chatting with the storage team at google
Notable bug fixes:
* rpk is effectively revamped and should be much less buggy from tuners
to logging, to namespacing, to containers and configuration re-write
* graceful recovery of an administration removing a directory manually
on /var/lib/redpanda/data/*