* Updated dependencies
Signed-off-by: Tomasz Baranski <tbaransk@redhat.com>
* Replace skopeo with containers API.
This commit removes dependence on skopeo (binary) and uses containers
API. By doing that we're able to opimize the use of storage (scratch)
space, storage I/O and download bandwith.
Signed-off-by: Tomasz Baranski <tbaransk@redhat.com>
* Fixing rebase - dependencies kerfuffle.
Signed-off-by: Tomasz Baranski <tbaransk@redhat.com>
* Handling docker-format images as well as OCI.
Signed-off-by: Tomasz Baranski <tbaransk@redhat.com>
* Fix for missing code-generator module.
Signed-off-by: Tomasz Baranski <tbaransk@redhat.com>
* Remove regex, image file in registry images are matched by a path
prefix.
Signed-off-by: Tomasz Baranski <tbaransk@redhat.com>
* Added nginx proxy in front of docker registry for a rate-limited access.
Signed-off-by: Tomasz Baranski <tbaransk@redhat.com>
* Combine tests in bad_webserver.go into a table
No functional change intended.
Signed-off-by: Maya Rashish <mrashish@redhat.com>
* Use a QCOW2 image for testing badserver.
This has the benefit of going through qemu-img for conversion,
which can introduce more bugs.
Signed-off-by: Maya Rashish <mrashish@redhat.com>
* Don't pass URLs to qemu-img if Accept-Ranges header isn't bytes.
Adapt unit tests so they don't also test this feature.
Add a handler to bad-webserver to do a very plain HTTP response
using a downloaded file from cdi-file-host. This one doesn't
come with builtin support for range requests.
Use this handler to test that CDI can still import images needing
conversion.
Signed-off-by: Maya Rashish <mrashish@redhat.com>
* Invert polarity and test for "none".
It's highly unlikely that a value other than bytes becomes a
valid choice, but let's not rule that out.
Signed-off-by: Maya Rashish <mrashish@redhat.com>
* update k8s deps to 1.18.6 and controller runtime to 0.6.2
Signed-off-by: Michael Henriksen <mhenriks@redhat.com>
* remove building code generators from docker image. This way the k8s ligray version only has to be updated in go.mod
Do more stuff in the bazel container. Faster and better interop
Fix unit tests
Signed-off-by: Michael Henriksen <mhenriks@redhat.com>
* make format
Signed-off-by: Michael Henriksen <mhenriks@redhat.com>
* remove unnecessary rsync
Signed-off-by: Michael Henriksen <mhenriks@redhat.com>
* redo code generator dep management
Signed-off-by: Michael Henriksen <mhenriks@redhat.com>
* builder uses go modules
Signed-off-by: Michael Henriksen <mhenriks@redhat.com>
* Move CRDS from apiextensions v1beta1 to v1.
Ensure that our code based schema validation matches the types in the api.
Signed-off-by: Alexander Wels <awels@redhat.com>
* Ran go mod tidy and vendor in attempt to see if we could use newer runtime controller, but our go version too old.
Addressed review comments.
Signed-off-by: Alexander Wels <awels@redhat.com>
* Addressed more review comments and fixed k8s-1.18 functional test failing.
Signed-off-by: Alexander Wels <awels@redhat.com>
* Remove categories 'all' from cluster scoped CRDs
Signed-off-by: Alexander Wels <awels@redhat.com>
* move upload.cdi.kubevirt.io API group to v1beta1
Signed-off-by: Michael Henriksen <mhenriks@redhat.com>
* move core api to v1beta1
Signed-off-by: Michael Henriksen <mhenriks@redhat.com>
* fix os-3.11 cluster sync and add functional tests for alpha api
Signed-off-by: Michael Henriksen <mhenriks@redhat.com>
* change more occurences of v1alpha1
Signed-off-by: Michael Henriksen <mhenriks@redhat.com>
* updates after rebase
Signed-off-by: Michael Henriksen <mhenriks@redhat.com>
* Correct check in badserver tests
Signed-off-by: Bartosz Rybacki <brybacki@redhat.com>
* Extend wait timeout for data volume success on flaky server
Signed-off-by: Bartosz Rybacki <brybacki@redhat.com>
* Do not use filesystem for state
Signed-off-by: Bartosz Rybacki <brybacki@redhat.com>
* Flaky server should only work on get.
Signed-off-by: Bartosz Rybacki <brybacki@redhat.com>
* Implement local web server to mimic flawed servers
http://cdi-bad-webserver.cdi/flaky/filename will return a 503
"unavailable" error to most requests.
http://cdi-bad-webserver.cdi/forbidden-HEAD/filename will return
a 403 "Forbidden" error on HEAD requests.
In both those cases, we validate that we still eventually succeed
in creating a datavolume.
Signed-off-by: Maya Rashish <mrashish@redhat.com>
* Add the imageio server to the manual test setup instructions.
Signed-off-by: Maya Rashish <mrashish@redhat.com>
* Add missing bad-webserver.yaml.in (forgot to git add)
Signed-off-by: Maya Rashish <mrashish@redhat.com>
* Make "flaky" bad-webserver reproducibly flaky.
9 failures followed by a success.
Switch to port 9090, os-3.11 disallows using privileged ports,
and we don't have a reason to prefer port 80.
Signed-off-by: Maya Rashish <mrashish@redhat.com>
* Provide polarion test IDs
Signed-off-by: Maya Rashish <mrashish@redhat.com>
* Make flaky test server fail less frequently.
The exponential backoff makes 10 retries very slow.
Signed-off-by: Maya Rashish <mrashish@redhat.com>
* Tolerate being installed on a namespace that isn't "cdi"
Assume that we are in the same namespace as cdi-file-host, and
use that namespace in the URL.
Signed-off-by: Maya Rashish <mrashish@redhat.com>
* Use AfterEach to cleanup datavolumes after test.
Suggested by Alexander Wels.
Signed-off-by: Maya Rashish <mrashish@redhat.com>
* Add loopback device container
Add code to create pv vg and lvm
Add ember csi lvm based provisioner.
Signed-off-by: Alexander Wels <awels@redhat.com>
* import/export
Signed-off-by: Alexander Wels <awels@redhat.com>
* Address PR comments.
Switch deployment to pod for loopback device
Sleep 1 instead of 60 to allow kubernetes easier way of killing pod.
Instead of embercsi master, grab one that works from known repo.
Signed-off-by: Alexander Wels <awels@redhat.com>
* Finish loopback pod instead of looping it.
Signed-off-by: Alexander Wels <awels@redhat.com>
* remove DOCKER_REPO from operator
Signed-off-by: Michael Henriksen <mhenriks@redhat.com>
* make generate and update CDI schema
Signed-off-by: Michael Henriksen <mhenriks@redhat.com>
* cut down operator permissions
Signed-off-by: Michael Henriksen <mhenriks@redhat.com>
* mage sure csv-generator is built but not included in operator image
Signed-off-by: Michael Henriksen <mhenriks@redhat.com>
* create aggregate cluster roles
Signed-off-by: Michael Henriksen <mhenriks@redhat.com>
* CDIConfig needs a ClusterRoleBinding so bind to system:authenticated for reading
Signed-off-by: Michael Henriksen <mhenriks@redhat.com>
* comment out problematic test
Signed-off-by: Michael Henriksen <mhenriks@redhat.com>
* operator needs upload.cdi.kubevirt.io permission
Signed-off-by: Michael Henriksen <mhenriks@redhat.com>
* revert commented out tests
Signed-off-by: Michael Henriksen <mhenriks@redhat.com>
The newer version of qemu-img (3.1.0) that ships in Fedora 30 has issues
with streaming conversions whereby it hangs during the conversion
process. Rather than try to force a downgrade of qemu when building
images instead, let's try reverting back to Fedora29 minimal which has
qemu-img version 3.0.0.
This issue has also exposed the fact that we aren't performing any
testing against streaming external images, we're only testing against
images that we've pulled in to our test infrastucture. So add a
functional test that builds a DataVolume directly from the cirros site.
set dv test back to expectin prlimit failure
* CDI OLM update support
* Download latest CDI OLM manifest from marketplace to set replaces field
* Prevent from genarting a csv version that already exist in the marketplace
* * Vendor the bellow packages in order to be able to fetch existing CSV
- github.com/operator-framework/operator-marketplace
- github.com/go-openapi/spec
- github.com/go-openapi/validate
- github.com/go-openapi/errors
- github.com/go-openapi/analysis
- github.com/go-openapi/strfmt
- github.com/go-openapi/loads
- github.com/go-openapi/runtime
* lint bug-fix
* CR fixes
* CR typo fixes
* * Split catalogsource/operatorsource/subscription manifests to k8s and os
* Update OLM documentation
* rename bundle-helper.go file
* * change CDI repository name in quay from cdi to cdi-operatorhub
* Update cdi-operator-olm.md
Added functional tests to ensure that malicious
qcow2 images are being rejected. I.e. the creation
of import datavolume should fail when converted to raw.
Change-Id: I3372105fdcd28b8e01811d7e6752f04f5dba9e0d
Signed-off-by: Daniel Erez <derez@redhat.com>
Signed-off-by: Fred Rolland <frolland@redhat.com>
Co-authored-by: Fred Rolland <frolland@redhat.com>
Looks like before some of imports used glog package that set all flags
under init() method, but after dependencies code stopped to run glog
init() method on container creation, so we need to set flags from klog
and parse flags in the right order.
- Generate OLM related manifests for CDI in _out/manifests/release/olm
OLM bundle:
- cdi CSV manifest
- cdi crd manifest
- cdi package manifest
- operatorsource manifest
- subscription manifest
- operatorgroup manifest
- Modify cdi-operator role not to be cluster-admin but more specific
- Move all final manifests to _out/manifests directory and update travis with new manifests location
- Provide API for vendoring CDI OLM manifests generation code
Note:
- OLM CSV update to be supported in a separate PR
- OLM bundle integration in travis is to be supported together with CSV update
