If NO_INSTALL_SEDNA not 'false', don't install Sedna, it's useful to
develop, the developer will install Sedna later by manually.
Signed-off-by: llhuii <liulinghui@huawei.com>
When added more than one cloud worker nodes, the master node is
unscheduled, this fix kubectl-taint-command.
Signed-off-by: llhuii <liulinghui@huawei.com>
When NUM_CLOUD_WORKERS >= 1, kind will taint master node with
key='node-role.kubernetes.io/master'.
By default, GM will be deployed at master node.
So make master node schedulable, and be able to run workloads, i.e. GM
here.
Signed-off-by: llhuii <liulinghui@huawei.com>
This avoid GM to be OOM killed temporarily.
TBD: Dive to analyze GM memory usage, which should be less than 128Mi.
Signed-off-by: llhuii <liulinghui@huawei.com>
supported protocols:
1. http(s)
2. s3 compatible with the env AWS_ACCESS_KEY_ID, AWS_SECRET_ACCESS_KEY,
AWS_ENDPOINT_URL
Since the image kfserving/storage-initializer is big(~341MB) and have
some limits with http server, the end image size is 54.7MB:
1. support to download more than one uri
2. smaller image size: use minio-client instead of boto3(less than
40MB), remove gcs/azure
3. remove the http-server content-type limits
4. add obs protocol
5. also extract .tar.gz/.zip for s3
6. fix pep8 style
Signed-off-by: llhuii <liulinghui@huawei.com>
1. retry to curl yamls
2. add 'SEDNA_ROOT' to install from existing source directory
3. force delete for the case deletion hangs
Signed-off-by: llhuii <liulinghui@huawei.com>
1. add all-in-one installation script. also support uninstall operation.
2. update the installation docs using the all-in-one script
Signed-off-by: llhuii <liulinghui@huawei.com>
JimmyYang20
llhuii
KubeEdge Bot