Fix 2 issues with config change detection:
- Objects received via Get() don't have kind and apiVersion set, while
those from Watch() do, leading to false positives.
- Compare the unmodified config (prior to applying defaults) to the
updated one from Watch().
Signed-off-by: Andy Goldstein <andy.goldstein@gmail.com>
- Adding in support for a new `download` subcommand of backup
- Adjusted signing to allow for multiple types
- Adding in git sha version during build more granular version
debugging
Signed-off-by: Justin Nauman <justin.r.nauman@gmail.com>
Delete all objects in backup "dir" when deleting a backup, instead of
hard-coding individual file names/types. This way, we'll be able to
delete log files and anything else we add without having to update our
deletion code.
Signed-off-by: Andy Goldstein <andy.goldstein@gmail.com>
Due to a change in glog, if you don't call flag.Parse, every log line
prints out 'ERROR: logging before flag.Parse'. This works around that
change. Ultimately we need to switch to a different logging library.
Signed-off-by: Andy Goldstein <andy.goldstein@gmail.com>
- Per discussion, there is no reason to deal
with the complexity of backwards compatibility
with the Namespace attribute on the Restore
domain.
- Also noticed there was an error on the
validation of the BackupController where
the message would actually just be the index
of the error and not the contents of the message
itself.
Signed-off-by: Justin Nauman <justin.r.nauman@gmail.com>
- Adding in additional test to ensure *Namespaces attributes
don't directly conflict logically with one another
- Additional PR changes around naming/typos
Signed-off-by: Justin Nauman <justin.r.nauman@gmail.com>
- Introduces similar Include/Exclude declaration on the Restore
resource and cli flags
- Kept support for legacy Namespaces attribute until it could be
deprecating. Defining both IncludeNamespaces and Namespaces results
in a validation error and the Restore will not be processed (shouldn't
be able to occur)
Signed-off-by: Justin Nauman <justin.r.nauman@gmail.com>
- Changed the default kubeconfig loading to utilize
the client-go's loader strategy. This allows users
to use the Ark client without having to explicitly
define a KUBECONFIG env var or argument.
This more closely resemebles how Kubectl works and users
are probably more used to while preserving the
current rules.
Signed-off-by: Justin Nauman <justin.r.nauman@gmail.com>
- Adding in paging support for the S3 and Snapshot
AWS integration.
As a testing note, you can add in a a MaxKeys to the S3
request as an easy way to ensure that paging is working
properly without having to creation over 1k backups.
Signed-off-by: Justin Nauman <justin.r.nauman@gmail.com>
