...
- NCSA Industry conference demo on Wednesday
- Workbench > Spark integration
- Zeppelin > Spark via Livy
- Spark > MongoDB using the NBI dataset
- Workbench > Spark integration
- SC17
- Workbench > HPC integration
- TERRA-REF image stitching on ROGER (TORQUE/PBS)
- Jupyter > Agave API > ROGER
- Jupyter > Agave API > Comet (Singularity)?
- Comet access
- XSEDE may be easiest
- Workbench > HPC integration
- Options:
- Deployment in commercial cloud ( multi-node cluster with storage)
- No one is happy with the Kubernetes deployment process on OpenStack
- kubeadm is promising
- Shared storage
- https://www.minio.io/?
- Review of OpenShift Origin
- Kubernetes, Swarm, Mesos have more traction
- Singularity and Shifter
- Security
- Authentication and authorization
- Oauth support in Workbench
- LDAP or Oauth scopes for authorization
- Q. What does it mean to share authentication/authorization with
- HPC cluster such as ROGER (NCSA LDAP)
- Spark cluster? (Kerberos or nothing, what about Livy)
- Container and filesystem users/permissions
- Mapping UID/GIDs into running containers
- Q. Root escalation in Docker
- Data sharing and permissions
- Controlling access to data on the filesystem but also in active database such as NBI mongo.
- Same problem with Spark (pulling data from Mongo into Spark)
- More sophisticated network configuration
- Both for Kubernetes
- Also with the cloud provider (e.g., OpenStack project)
- Authentication and authorization
- Monitoring/LMA
- Addons https://kubernetes.io/docs/concepts/cluster-administration/addons/
- Prometheus?
- Nagios/NRPE is a stopgap
- Production maintenance
- Upgrading the beta instance
- Redeploying the beta?
- /var/lib/docker volume scaling
- Fixing inactiveTimeout for inactive accounts
- ETK/earthcube insteancesinstances
- Cloud9 wily (and other Wily) upgrades; use Cloud9 all
- Process for removing old specs if in use.
- Clowder/Workbench plugin?
- Evaluate FRDR
- Deployment in commercial cloud ( multi-node cluster with storage)
...