Releases: fluxcd/flux
Flux 1.8.0
This release includes a change to how image registries are scanned for metadata, which should reduce the amount of polling, while being sensitive to image metadata that changes frequently, as well as respecting throttling.
Fixes
- Better chance of a graceful shutdown on signals weaveworks/flux#1438
- Take more notice of possible errors weaveworks/flux#1432 and weaveworks/flux#1433
- Report the problematic string when failing to parse an image ref weaveworks/flux#1407
Improvements
- Apply CustomResourceDefinition manifests ahead of (most) other kinds of resource, since there will likely be other things that depend on the definition (e.g., the custom resources themselves) weaveworks/flux#1429
- Add
--git-timeout
flag for setting the default timeout for git operations (useful e.g., if you knowgit clone
will take a long time) weaveworks/flux#1416 fluxctl list-controllers
now has an aliasfluxctl list-workloads
weaveworks/flux#1425- Adapt the sampling rate for image metadata, and back off when throttled weaveworks/flux#1354
- The detailed rollout status of workloads is now reported in the API (NB this is not yet used in the command-line tool) weaveworks/flux#1380
Thanks
A warm thank-you to @AugustasV, @MansM, @Morriz, @MrYadro, @Timer, @aaron-trout, @bhavin192, @brandon-bethke-neudesic, @brantb, @bzon, @dbluxo, @dholbach, @dlespiau, @endrec, @hiddeco, @justdavid, @justinbarrick, @kozejonaz, @lelenanam, @leoblanc, @marcemq, @marcusolsson, @mellena1, @mt-inside, @ncabatoff, @pcfens, @rade, @rndstr, @sc250024, @sfrique, @skurtzemann, @squaremo, @stefanprodan, @stephenmoloney, @timthelion, @tlvu, @whereismyjetpack, @white-hat, @wstrange for your contributions.
Helm operator 0.3.0
This release adds dependency handling to the Helm operator.
NB The helm operator will now update dependencies for charts by
default, which means you no longer need to vendor them. You can
switch this behaviour off with the flag --update-chart-deps=false
.
Bug fixes
- Improve chance of graceful shutdown
weaveworks/flux#1439
and
weaveworks/flux#1438
Improvements
- The operator now runs
helm dep build
for charts before installing
or upgrading releases. This will use a lockfile if present, and
update the dependencies according torequirements.yaml
otherwise
weaveworks/flux#1450 - A new flag
--git-timeout
controls how long the Helm operator will
allow for git operations
weaveworks/flux#1416 - The Helm operator image now includes the Helm command-line client,
which makes it easier to troubleshoot problems usingkubectl exec
(as part of
weaveworks/flux#1450)
Many thanks to everyone who contributed!
Flux 1.7.1
This is a patch release, mainly to include the fix for initContainer
images (#1372).
Fixes
- Include initContainers when scanning for images to fetch metadata
for, e..g, so there will be "available image" rows for the
initContainer influxctl list-images
weaveworks/flux#1372 - Turn memcached's logging verbosity down, in the example deployment
YAMLs weaveworks/flux#1369 - Remove mention of an archaic
fluxctl
command from help text
weaveworks/flux#1389
Thanks
Thanks for fixes go to @alanjcastonguay, @dholbach, and @squaremo.
Flux 1.7.0
This release has a soupçon of bug fixes. It gets a minor version bump, because it introduces a new flag, --listen-metrics
.
Fixes
- Updates to workloads using initContainers can now succeed weaveworks/flux#1351
- Port forwarding to GCP (and possibly others) works as intended weaveworks/flux#1334
- No longer falls over if the directory given as
--git-path
doesn't exist weaveworks/flux#1341 fluxctl
doesn't try to connect to the cluster when just reporting its version weaveworks/flux#1332- Metadata for unusable images (e.g., those for the wrong architecture) are now correctly recorded, so that they don't get fetched continually weaveworks/flux#1304
Improvements
- Prometheus metrics can be exposed on a port different from that of the flux API, using the flag
--listen-metrics
weaveworks/flux#1325
Thanks
Thank you to the following for contributions (along with anyone I've missed): @ariefrahmansyah, @brantb, @casibbald, @davidkarlsen, @dholbach, @hiddeco, @justinbarrick, @kozejonaz, @lelenanam, @petervandenabeele, @rade, @rndstr, @squaremo, @stefanprodan, @the-fine.
Helm operator 0.2.1
This is a patch release that allows helm-op to recover from a failed release install. If a chart is broken, Tiller will reserve the name and mark the release as failed. If at a later time the chart is fixed, helm-op can't install it anymore because the release name is in use. Purging the release after each failed install allows helm-op to keep retrying the install.
- Purge release if install fails weaveworks/flux#1344
Flux 1.6.0
This release improves existing features, and has some new goodies like
regexp tag filtering and multiple sync paths. Have fun!
We also have a new contributing guide.
Fixes
- Update example manifests to Kubernetes 1.9+ API versions
weaveworks/flux#1322 - Operate with more restricted RBAC permissions
weaveworks/flux#1298 - Verify baked-in host keys (against known good fingerprints) during
build
weaveworks/flux#1283 - Support authentication for GKE, AWS, etc., when
fluxctl
does
automatic port forwarding
weaveworks/flux#1284 - Respect tag filters in
fluxctl release ...
, unless--force
is
given
weaveworks/flux#1270
Improvements
- Cope with
':'
characters in resource names
weaveworks/flux#1282 - Accept multiple
--git-path
arguments; sync (and update) files in
all the paths given
weaveworks/flux#1297 - Use image pull secrets attached to service accounts, as well as
those attached to workloads themselves
weaveworks/flux#1291 - You can now filter images using regular expressions (in addition to
semantic version ranges, and glob patterns)
weaveworks/flux#1292
Thanks
Thank you to the following for contributions: @Alien2150,
@ariefrahmansyah, @brandon-bethke-neudesic, @bzon, @dholbach,
@dkerwin, @hartmut-pq, @hiddeco, @justinbarrick, @petervandenabeele,
@nicolerenee, @rndstr, @squaremo, @stefanprodan, @stephenmoloney.
Helm operator 0.2.0
In large part this release simplifies and improves the Helm operator
machinery, without changing its effect.
This release drops the -alpha
suffix, but remains <1.0 and should
(still) be considered unready for production use.
- Use the same git implementation as fluxd, fixing a number of
problems with SSH known_hosts and git URLs and so on
weaveworks/flux#1240 - Always check that a chart release will be a change, before releasing
weaveworks/flux#1254 - Add validation to the FluxHelmRelease custom resource definition,
giving the kind the short namefhr
weaveworks/flux#1253 - Detect chart release differences more reliably
weaveworks/flux#1272 - Check for more recent versions and report in logs when out of date
weaveworks/flux#1276
See getting started with
Helm
and the Helm chart
instructions
for information on installing the Flux with the Helm operator.
Flux 1.5.0
This release adds semver image filters, makes it easier to use
fluxctl
securely, and has an experimental interactive mode for
fluxctl release
. It also fixes some long-standing problems with
image metadata DB, including no longer being bamboozled by Windows
images.
Fixes
- Read the fallback image credentials every time, so they can be
updated. This makes it feasible to mount them from a ConfigMap, or
update them with a sidecar
weaveworks/flux#1230 - Take some measures to prevent spurious image updates caused by bugs
in image metadata fetching:- Sort images with zero timestamps correctly
weaveworks/flux#1247 - Skip any updates where there's suspicious-looking image metadata
weaveworks/flux#1249
(then weaveworks/flux#1250) - Fix the bug that resulted in zero timestamps in the first place
weaveworks/flux#1251
- Sort images with zero timestamps correctly
- Respect
'false'
value for automation annotation
weaveworks/flux#1264 - Cope with images that have a Windows (or other) flavour, by omitting
the unsupported image rather than failing entirely
weaveworks/flux#1265
Improvements
fluxctl
will now transparently port-forward to the Flux pod,
making it easier to connect securely to the Flux API
weaveworks/flux#1212fluxctl release
gained an experimental flag--interactive
that
lets you toggle each image update on or off, then apply exactly the
updates you have chosen
weaveworks/flux#1231- Flux can now report and update
initContainers
, and a wider variety
of Helm charts (as used inFluxHelmRelease
resources)
weaveworks/flux#1258 - You can use semver (Semantic Versioning) filters
for automation, rather than having to rely on glob patterns
weaveworks/flux#1266
Thanks
Thanks to @ariefrahmansyah, @chy168, @cliveseldon, @davidkarlsen,
@dholbach, @errordeveloper, @geofflamrock, @grantbachman, @grimesjm,
@hiddeco, @jlewi, @JoeyX-u, @justinbarrick, @konfiot, @malvex,
@marccampbell, @marctc, @mt-inside, @mwhittington21, @ncabatoff,
@rade, @rndstr, @squaremo, @srikantheee84, @stefanprodan,
@stephenmoloney, @thejaysmith (and anyone I've missed!) for their
contributions.
Helm operator release 0.1.1-alpha
This patch release bundles a number of contributed fixes, to make the Helm operator work in a broader variety of circumstances (i.e., not just as a demo). The Helm operator is still considered "alpha".
- Support using TLS connections to Tiller
weaveworks/flux#1200 - Avoid continual, spurious installs in newer Kubernetes
weaveworks/flux#1193 - Make it easier to override SSH config (and
known_hosts
)
weaveworks/flux#1188 - Annotate resources created by a Helm release with the name of the
FluxHelmRelease custom resource, so they can be linked
weaveworks/flux#1134 - Purge release when FluxHelmRelease is deleted, so restoring the
resource can succeed
weaveworks/flux#1106 - Correct permissions on baked-in SSH config
weaveworks/flux#1098 - Test coverage for releasesync package
weaveworks/flux#1089).
It is now possible to install Flux and the Helm operator using the helm chart in this
repository.
Flux release 1.4.2
This release includes a number of usability improvements, the majority
of which were suggested or contributed by community members. Thanks
everyone!
Fixes
- Don't output fluxd usage text twice
weaveworks/flux#1183 - Allow dots in resource IDs; e.g.,
default:deployment/foo.db
, which
is closer to what Kubernetes allows
weaveworks/flux#1197 - Log more about why git mirroring fails
weaveworks/flux#1171
Improvements
- Interpret FluxHelmRelease resources that specify multiple images to
use in a chart
weaveworks/flux#1175
(and several PRs that can be tracked down from there) - Add an experimental flag for restricting the view fluxd has of the
cluster, reducing Kubernetes API usage:--k8s-namespace-whitelist
weaveworks/flux#1184 - Share more image layers between quay.io/weaveworks/flux and
quay.io/weaveworks/helm-operator images
weaveworks/flux#1192 - Apply resources in "dependency order" so that e.g., namespaces are
created before things in the namespaces
weaveworks/flux#1117