* test: remove group skips
* test: inline previously extended testcases that broke with added sort
Some tests had a sort added so they'd pass in cloud, vanilla, and OSS.
This also broke some extended testcases here in OSS since pushdown rules
no longer match the plan.
This diff "inlines" the testcase body for each of these so we get our
coverage back, but leaves a FIXME to note that it would be better if we
could somehow thread the needle and retain the extension.
* fix: allow backup of all buckets
* Revert "fix: allow backup of all buckets"
This reverts commit 256ec2f591a36b946f5c46014f378b2b76c9a8de.
* fix: remove 20 item limit in several of the data stores
* test: add a test for 20 item limit
* feat(security): set SameSite=strict on session cookie
Use SameSite=Strict as a hardening measure against cross-origin attacks.
While browsers have been moving to default to SameSite=Lax, explicitly
setting SameSite ensures that all browsers enforce it consistently.
While 'lax' is a reasonable hardening choice, the cookie is only
required for requests to '/api/...' and we don't expect 3rd party links
into '/api/...', so this stricter setting should be safe in terms of
usability. Furthermore, while our GET APIs are not state-changing, using
'strict' future-proofs us in case we add a state-changing GET API ('lax'
allows cross-origin 'GET' requests for increased usability for read-only
requests).
Also add a comment to SetCORS() lack of Access-Control-Allow-Credentials
as a reminder that its omission is intentional for defense in depth on
when to attach the cookie to a request.
* chore: mention that Lax sends the cookie with other safe HTTP methods
* chore: add protoc-gen script to releng
* chore: break cross-container-tag into separate variable
* fix: call GNUMakefile "generate-sources" target instead
This also does a better job at mounting the root directory
in the docker container.
This fixes an occurrence of a loop variable being captured in a
parallel test. With the previous code, only the last test case is
actually exercised. To work around this problem, we create a local
copy of the range variable before the parallel test, as suggested in
the Go documentation for the `testing` package:
https://pkg.go.dev/testing#hdr-Subtests_and_Sub_benchmarks
Issue was found automatically using the `loopvarcapture` linter.
* feat: add the ability to replicate based on bucket name rather than bucket id.
- This adds compatibility with 1.x replication targets
* fix: improve error checking and add tests
* fix: add additional constraint to replications table
* fix: use OR not AND for constraint
* feat: delete invalid replications on downgrade
* fix: should be less than 2.4
* test: add test around down migration and cleanup migration code
* fix: use nil instead of platform.ID(1) for better consistency
* fix: fix tests
* fix: fix tests
* chore: fix build commands for new cross-builder
* fix: use xcc to determine compiler
* fix: use correct compiler for darwin tests
* fix: update perftest to 22.04 image
* fix: fix find dbrps, make bucket default if not overridden
* fix: allow filtering of virtual DBRPs, filter in shard mapper
* fix: update tests to mock for virtual filter for shards and update server test
* feat: Add the instance owner concept
* feat: improve handling of instance resource type
* fix: remove relic of cloud/oss split
* feat: Add an instance id. This gets set during onboarding and can be used to uniquely identify that instance.
* chore: Revert "feat: Add an instance id. This gets set during onboarding and can be used to uniquely identify that instance."
This reverts commit 6a712df61e.
We decided to stick with the "singleton" 1 id for now, as there is not a clear need for unique identifiers.
* chore: Update PULL_REQUEST_TEMPLATE.md
This is part of an initiative to disseminate PR-specific information for all PR stakeholders. In addition to reviewers, this information is needed by product and docs.
* chore: Update PULL_REQUEST_TEMPLATE.md
* chore: Apply suggestions from code review
Co-authored-by: Sam Arnold <sarnold@influxdata.com>
Co-authored-by: Jeffrey Smith II <jsmith@influxdata.com>
Co-authored-by: Sam Arnold <sarnold@influxdata.com>
Whilst working on cloud labels, I noticed that the label storage test
isn't correctly capturing the testScenario. This is because t.Parallel()
is in use, so only the "delete" scenario will only ever be tested.
The following changes fix that, by capturing the testScenario in a local
variable so that the test function can use that correctly.
As a drive-by, I fixed the error referencing the wrong expectation.
Nathaniel Cook
Jason Stirnaman
Jeffrey Smith II
Owen Nelson
Christopher M. Wolff
Brandon Pfeifer
Jamie Strandboge
davidby-influx
Cemre Mengu
Abirdcfly
Dane Strandboge
Jonathan A. Sternberg
Renato Costa
Sean Brickley
Andrew Lee
Ikko Ashimine
Sam Dillard
Simon Richardson
Sunil Kartikey