This uses SQLite-backed informers provided by Lasso with https://github.com/rancher/lasso/pull/65 to implement Steve API (/v1/) functionality.
This new functionality is available behind a feature flag to be specified at Steve startup
See https://confluence.suse.com/pages/viewpage.action?pageId=1359086083
Co-authored-by: Ricardo Weir <ricardo.weir@suse.com>
Co-authored-by: Michael Bolot <michael.bolot@suse.com>
Co-authored-by: Silvio Moioli <silvio@moioli.net>
Signed-off-by: Silvio Moioli <silvio@moioli.net>
ByNames could previously return a nil value and a nil error. This caused
issues when other parts of the application
(pkg/stores/partition/parallel.go) tried to use the result. Now this
will return an empty list on the error condition, instead of nil
Remove unnecessary items from the `access` slice in one of the unit
tests. Each item in `access` corresponds to one request in `apiOps`, so
they should have the same number of elements.
Add a nested store to the proxy store to strip non-Kubernetes fields
from the object being updated.
The steve formatter and proxy store adds fields to objects when it
outputs them to the client, for usability by the UI. It adds the
object's fields[1], relationships to other objects[2], a summary of the
object's state[3], and additional information in the conditions[4].
These fields are not native to Kubernetes, so when a client submits the
object back as an update, Kubernetes reports a warning that they are
unrecognized. This change ensures the extra fields are removed before
submitting the update.
[1] bf2e9655f5/pkg/stores/proxy/proxy_store.go (L189)
[2] bf2e9655f5/pkg/resources/common/formatter.go (L106)
[3] bf2e9655f5/pkg/resources/common/formatter.go (L100)
[4] bf2e9655f5/pkg/resources/common/formatter.go (L108)
Add a new query parameter to filter resources by their namespace or
their namespace's project. This parameter is separate from the existing
`filter` parameter.
Filter by a comma-separated list of projects and/or namespaces with:
?projectsornamespaces=p1,n1,n2
The result can be negated with the ! operator:
?projectsornamespaces!=p1,n1,n2
This change adds support for excluding results using the != operator.
Example to exclude all results with name "example":
?filter=metadata.name!=example
Include all results from namespace "example" but exclude those with name
"example":
?filter=metadata.namespace=example&metadata.name!=example
Exclude results with name "foo" OR exclude results with name "bar"
(effectively includes results of both types):
?filter=metadata.name!=foo,metadata.name!=bar
Currently, multiple filters can be appended on the query string, and
each subsequent filter is ANDed with the set. Items that pass through
the filter set must match every filter in the set.
This change adds support for OR filters. A single filter key can specify
multiple filters, separated by ','. An item that passes this filter can
match any filter in the set.
For example, this filter matches items that have either "name" or
"namespace" that match "example":
?filter=metadata.name=example,metadata.namespace=example
This filter matches items that have "name" that matches either "foo" or
"bar":
?filter=metadata.name=foo,metadata.name=bar
Specifying more than one filter key in the query still ANDs each inner
filter set together. This set of filters can match either a name of
"foo" or "bar", but must in all cases match namespace "abc":
?filter=metadata.name=foo,metadata.name=bar&filter=metadata.namespace=abc
Without this change, the cache is checked if a revision is specified,
but the revision is ignored if the value is not found in the cache. This
is a problem if by chance sequential paginated requests land on separate
pods, because a different version number may be fetched on the sequent
request and cause the caches to be inconsistent with one another. This
change ensures that `revision` is passed as `resourceVersion` to the
Kubernetes request, since `revision` has no meaning to Kubernetes.
The default chunk size in the partition was set to 100000. It could be
overridden as a larger or smaller number, but not disabled altogether.
This change adds the ability for users to explicitly opt out of the
limit by specifying a negative number or zero. The default behavior is
the same.
Extend the sorting functionality in the partition store to support
primary and secondary sorting criteria. Sorting parameters are specified
by a single 'sort' query and comma separated. Example:
Sort by name and creation time:
GET /v1/secrets?sort=metadata.name,metadata.creationTimestamp
Reverse sort by name, normal sort by creation time:
GET /v1/secrets?sort=-metadata.name,metadata.creationTimestamp
Normal sort by name, reverse sort by creation time:
GET /v1/secrets?sort=metadata.name,-metadata.creationTimestamp
Cache filtered, sorted results for fast subsequent page retrieval.
Requests for cached queries need to include the list revision number
along with other queries. If no specific revision is requested, a new
fetch is done in order to get the latest revision. The revision is
included in the list response.
Example first request:
GET /v1/secrets?pagesize=10
Example subsequent page request:
GET /v1/secrets?pagesize=10&page=1&revision=107740
Extend the partition store to parse page and pagesize parameters and
return a subset of list results. The total number of pages is included
in the list response.
Request an initial page:
GET /v1/secrets?pagesize=10
or
GET /v1/secrets?pagesize=10&page=1
Request subsequent pages, or arbitrary pages:
GET /v1/secrets?pagesize=10&page=37
If a page number is out of bounds, an empty list is returned.
Extend the partition store to parse the "sort" query parameter as a
sorting condition. Dot notation is used to denote the object field.
Preceding the key with "-" denotes descending (reverse) order.
Example sorting by name:
GET /v1/secrets?sort=metadata.name
Reverse sorting by name:
GET /v1/secrets?sort=-metadata.name
All values are converted to strings and sorted lexicographically.
Extend the partition store to parse query params as list filters. Filter
keys use dot notation to denote the subfield of an object to filter on.
Example:
GET /v1/secrets?filter=metadata.name=foo
Filters are ANDed together, so an object must match all filters to be
included in the list. Example:
GET /v1/secrets?filter=metadata.name=foo&filter=metadata.namespace=bar
Arrays are searched for matching items. If any item in the array
matches, the item is included in the list. Example:
GET /v1/pods?filter=spec.containers.image=alpine
Filtering and sorting needs to operate on unstructured data. It also
needs to operate after the parallel partitioner, higher in the store
stack. This means that the proxy Store needs to return raw, unstructured
data up to the partitioner. This change moves all conversions from
unstructured Kubernetes types to apiserver objects up from the proxy
store into the partitioner.
The use of AddNamespaceConstraint was removed in e35b8304 of
rancher/rancher, so there is no possibility of there being a namespace
constraint in the request context. Remove the unused function and the
unused codepath from the rbac store.
By default, a watch times out after 30 minutes. For debugging purposes,
it's convenient if this can be decreased. Add an environment variable
CATTLE_WATCH_TIMEOUT_SECONDS to enable setting the timeout in seconds.
Add debug logs and send websocket messages when the watch is closed
unexpectedly.
In addition to being helpful for debugging, the dashboard specifically
looks for a `resource.error` event containing the string "too old" in
order to trigger the watch to be resynced with a refreshed revision
number. Without this error returned, the dashboard will only see
`resource.stop` events and never change its behavior, continuing to try
to restart the watch with an incorrect resource version.
