Releases: minio/minio
Bugfix and Performance improvement release
Highlights
- Due to major reduction in overall internode connection usage, upto 25% performance improvement
is see for a lot of small objects ranging from 4k to 1MiB - visible via faster response times on NVMe
setups. - Performance improvement for SSE-KMS/SSE-C uploads of multipart in nature, to avoid md5sum
entirely when it is not requested to be computed by client.
What's Changed
- reject expired STS credentials early without decoding sessionToken by @harshavardhana in #19072
- Convert more peer peer REST calls by @klauspost in #19004
- simplify broker healthcheck by following kafka guidelines by @harshavardhana in #19082
- Graph cluster and bucket replication proxied requests by @shtripat in #19078
- allow renames() for inlined writes without data-dir by @harshavardhana in #18801
- do not block on send channels under high load by @harshavardhana in #19090
- site-repl: Fix endpoint in the error with unexpected deployment-id by @vadmeste in #19086
- Add missing yaml syntax highlighting to docs/metrics/prometheus/README.md by @schmittey in #19087
- site replication: Disallow removal of site-replicator account by @poornas in #19092
- Fix grid reconnection deadlock by @klauspost in #19101
- Add array recycling safety by @klauspost in #19103
- Add DeleteAll with expiry days non zero value only by @shtripat in #19095
- use typos instead of codespell by @harshavardhana in #19088
- turn-off md5sum for SSE-KMS/SSE-C as optimization for multipart by @harshavardhana in #19106
- Add audit for folder excess by @vadmeste in #19109
- s3: CopyObject to disallow invalid dest object names by @vadmeste in #19110
- ilm: Update action count only on success by @krisis in #19093
- fix: allow diskInfo at storageRPC to be cached by @harshavardhana in #19112
- move timedValue -> cachevalue package by @harshavardhana in #19114
- fix: re-use endpoint strings to avoid allocation during audit by @harshavardhana in #19116
- allow all disk full errors to be handled by @harshavardhana in #19117
New Contributors
- @schmittey made their first contribution in #19087
Full Changelog: RELEASE.2024-02-17T01-15-57Z...RELEASE.2024-02-24T17-11-14Z
Bugfix release
What's Changed
- remove rename2 entirely, avoids the risk of moving data by @harshavardhana in #19058
- Refresh tier config periodically by @krisis in #19049
- sr: add validation to disallow updating bandwidth limit on self by @poornas in #19062
- log: Add logger.Event to send to console and other logger targets by @vadmeste in #19060
- converge clusterRead health into cluster health by @harshavardhana in #19063
- Fix typo in jwt skewed date/time error by @vadmeste in #19066
- fix: found races in accessing globalLocalDrives by @harshavardhana in #19069
Full Changelog: RELEASE.2024-02-14T21-36-02Z...RELEASE.2024-02-17T01-15-57Z
Bugfix release
What's Changed
- fix: remove unnecessary panic in iam-store by @donatello in #19050
- fix sr tests to capture all server logs by @poornas in #19051
- fix: race in deleting objects during batch expiry by @Praveenrajmani in #19054
- send correct objectname in audit events for DeleteAll ILM by @harshavardhana in #19053
- upgrade all deps by @harshavardhana in #19041
- fix: reject service account access key same as root credentials by @harshavardhana in #19055
- fix: revert allow offline disks on fresh start by @harshavardhana in #19052
- add customer specific hotfixes to 'registry.min.dev' by @harshavardhana in #19057
Full Changelog: RELEASE.2024-02-13T15-35-11Z...RELEASE.2024-02-14T21-36-02Z
Bugfix release
What's Changed
- FIx unexpected behavior when creating service account by @taran-p in #19036
- sts: Add test for DurationSeconds condition by @vadmeste in #19044
- add missing handler for reloading site replication config on peers by @harshavardhana in #19042
- fix: update batch replication stats for snowball uploads by @Praveenrajmani in #19045
- Send a bucket notification event on DeleteObject() for non-existing object by @Praveenrajmani in #19037
- fix incorrect disk io stats in k8s environment by @anjalshireesh in #19016
Full Changelog: RELEASE.2024-02-12T21-02-27Z...RELEASE.2024-02-13T15-35-11Z
Bugfix Release
What's Changed
- avoid excessive logging for objects that do not exist by @harshavardhana in #19030
- Fix panic in tagging request proxying by @poornas in #19032
- do not have to use the same distributionAlgo as first pool by @harshavardhana in #19031
- fix: allow configuring excess versions alerting by @harshavardhana in #19028
- preserve conflicting objects when parent object is being deleted by @harshavardhana in #19034
- Convert service account add/update expiration to cond values by @vadmeste in #19024
- relax pre-emptive GetBucketInfo() for multi-object delete by @harshavardhana in #19035
- honor replaced disk properly by updating globalLocalDrives by @harshavardhana in #19038
- metrics: fix typo in namespace for proxy tagging metric by @poornas in #19039
Full Changelog: RELEASE.2024-02-09T21-25-16Z...RELEASE.2024-02-12T21-02-27Z
Bugfix release
What's Changed
- do not block iam.store registration by @harshavardhana in #18999
- listing must return WalkDir() errors first by @harshavardhana in #19006
- Add GetBucketInfo toStorageErr conversion by @klauspost in #19005
- fix: skip policy usage validation for cache update by @donatello in #19008
- Update IAM access manager plugin demo by @donatello in #19007
- Fix blocked streams blocking reconnects by @klauspost in #19017
- optimize startup sequence performance by @harshavardhana in #19009
- Fix shared top locks client by @klauspost in #19018
- fix: dangling objects honor parityBlocks instead of dataBlocks by @harshavardhana in #19019
- remove unnecessary metrics in 'mc admin info' output by @harshavardhana in #19020
- Add extra disconnect safety by @klauspost in #19022
- introduce reader deadlines for net.Conn by @harshavardhana in #19023
Full Changelog: RELEASE.2024-02-06T21-36-22Z...RELEASE.2024-02-09T21-25-16Z
Bugfix release
Highlights
- Fixes a second memory leak observed in ReadVersion without data that uses websockets layer.
- Fixes a crash while running
mc admin trace
on a distributed setup. - Fixes TCP socket hangs by adding short deadlines.
What's Changed
- avoid excessive logs when peer is down by @harshavardhana in #18969
- remove unused DiskMTime by @harshavardhana in #18965
- Add a defensive code to ignore checking parts with transitioned objects by @vadmeste in #18973
- Disable Rename2 metrics on non-linux by @klauspost in #18970
- sts: Add support of AssumeRoleWithWebIdentity and DurationSeconds by @vadmeste in #18835
- do not count context canceled as timeout errors by @harshavardhana in #18975
- calculate correct quorum check for metadata updates on object by @harshavardhana in #18979
- metrics: add replication metrics on proxied requests by @poornas in #18957
- Add missing gob register for map[string]string{} by @harshavardhana in #18974
- protect tracker copy properly to avoid race by @harshavardhana in #18984
- Fix tracing send on closed channel by @klauspost in #18982
- Add websocket TCP write timeouts by @klauspost in #18988
- Fix second muxclient context leak by @klauspost in #18987
- Enable expired-object-all-versions by @shtripat in #18954
Full Changelog: RELEASE.2024-02-04T22-36-13Z...RELEASE.2024-02-06T21-36-22Z
Bugfix release
Highlights
-
Fixes a regression in service accounts/access keys for the root-user stopped working in the last release,
a side-affect of a security bug-fix has been addressed. -
Fixes for rename2() usage in decommission and rebalance on versioned buckets with objects having
more than a version. this can lead to an inconsistent state() on the object requiring manual intervention.
If you are planning to rebalance or decommission on a cluster you are advised to upgrade to this
release - #18964 -
Fixes for memory leaks observed in the new websockets implementations, causing a gradual build-up
of memory in GiBs on busy clusters with heavy READ operations. -
Optimizations for internode communication for variety of APIs ranging from IAM, Bucket Metadata.
reducing internode connections to bare-minimum, there will be some more optimizations coming in this
area for deploying MinIO at > 300 node clusters.
What's Changed
- move a collection of peer APIs to websockets by @harshavardhana in #18936
- Improve object reuse for grid messages by @klauspost in #18940
- batch: Fix a typo while validating smallerThan field by @vadmeste in #18942
- fix:
null
inline policy handling for access keys by @donatello in #18945 - log: Use error log type instead of Application/MinIO type by @vadmeste in #18930
- Fix some leftover issues from PR 18936 by @fwessels in #18946
- simplify deadlineWriter, re-use WithDeadline by @harshavardhana in #18948
- Fix ineffective recycling by @klauspost in #18952
- deprecate disk tokens, instead rely on deadlines and active monitoring by @harshavardhana in #18947
- move Make,Delete,Head,Heal bucket calls to websockets by @harshavardhana in #18951
- Fix mux client memory leak by @klauspost in #18956
- disconnected returns, an unexpected error to List() returning 500s by @harshavardhana in #18959
- xl: Disable rename2 in decommissioning/rebalance by @vadmeste in #18964
- Fix typo in api-router.go by @fwessels in #18955
- Add more advanced cases for dangling by @harshavardhana in #18968
Full Changelog: RELEASE.2024-01-31T20-20-33Z...RELEASE.2024-02-04T22-36-13Z
Security Bugfix Release
Highlights
- This release fixes a security issue related to service accounts and their permissions for more details
- Performance improvement for large clusters with high IOPs requirements.
What's Changed
- fix: bucket metric of data usage for replication info will be ignore by @jiuker in #18895
- fix metrics panic in node metrics endpoint by @poornas in #18894
- Update list.md by @dvaldivia in #18907
- allow configuring maximum idle connections per host by @harshavardhana in #18908
- enable xattr capture by default by @harshavardhana in #18911
- reuse transports for callhome and remote tgt validation by @poornas in #18912
- exclude veeam virtual objects from replication by @poornas in #18918
- Add cgroup v2 support for memory limit by @vadmeste in #18905
- remove checkBucketExist check entirely to avoid fan-out calls by @harshavardhana in #18917
- Improve tracing & notification scalability by @klauspost in #18903
- use all the available nr_requests for NVMe by @harshavardhana in #18920
- add total usable capacity, free and used to DataUsageInfo() by @harshavardhana in #18921
- remove all the frivolous logs, that may or may not be actionable by @harshavardhana in #18922
- Update service file version in makefile by @donatello in #18925
- Correct small typo by @fwessels in #18923
- fix: permission checks for editing access keys by @donatello in #18928
- Keep ServiceV1 admin stop/restart API and mark as deprecated by @vadmeste in #18932
Full Changelog: RELEASE.2024-01-29T03-56-32Z...RELEASE.2024-01-31T20-20-33Z
Bugfix Release
Full Changelog: RELEASE.2024-01-28T22-35-53Z...RELEASE.2024-01-29T03-56-32Z
Highlights
- Fixes an ugly issue with runtime debug stack excessive logging regression from
RELEASE.2024-01-28T22-35-53Z, users who upgraded to RELEASE.2024-01-28T22-35-53Z
are advised to upgrade immediately.