July 27, 2016 Nebula OpenStack user's meeting These meetings will be regularly scheduled every 2 months and will last an hour. They will consist mostly of user presentations and conversations with about 10 minutes of technical presentation by the ITS staff. Doug Fein's technical presentation Nebula increasted from 18 nodes (8 storage and 10 compute) to 65 nodes (30 storage and 35 compute). Over the last month there were 1 million virtual cpu hours used on Nebula. Over 300 VMs were active at a time. 11 Billion GB hours, 1.3 PB of data allocated on the system (but less than 15% of it was used). There are about 33 active projects plus other projects that are starting up. There are more than 250 active users on Nebula. All nodes have dual 10 GigE interconnect. Chris Lindsey is the primary administrator of Nebula, Shaq is the junior administrator. Shaq is handling account and group related tickets. Backend storage is moving to glusterFS. Working on implementing live migration of VM images from node to node to prevent downtime. Testing Swift object store for LSST. It will be rebuilt on 8/15. After that it can grow to other users. OpenStack Kilo does not have good object quotas. Have to decide whether object store is replicated or not. About 58TB on each storage node + compression makes it closer to 68TB per node. Shared nova storage Cinder volume mounts gluster Downtime of < 10 minutes for most instances. Move node by node into live migration. Hoping to start next week. Best to be done before start of academic year. Possible to have a couple of bare metal nodes? Chameleon - bare level tweaking Looking into LDAP integration, Native Docker control. December upgrade to Mitaka - take a 1 or 2 day outage. Possibly during the Christmas break. Docker, LDAP, Mitaka - 3 primary areas being worked on Sahara? Need to determine user requirements for IP addresses. What are project's needs for Swift. The intention is that users start with an introductory allocation, when their needs grow past that limit it is expected that they will purchase equipment (storage and compute nodes) to increase their quotas. 20 VMs, 40 virtual cpus, 5TB of storage is teh minimum allocation. Major users are working on buy in. Peter will collaborate with Jeff T to develop getting started documentation. We will gather testimonials from Nebula users - Rob is working on this. David Raila of NDS labs spoke about his group's use of Nebula Research data tool development Platform as a service on Kubernetes Hard to set up the tools - they are working to make a turnkey workbench. They are using Core OS - cluster operating system, updates itself They use Ansible to set thigs up. Nothing is installed on the endpoint. Docker is efficient, small and fast. Brunner is interested in a Kubernetes image. Rob Kooper of ISDA presented slides about his group's use of Nebula They have scripts that use OpenStack's API They also use Puppet Interested in a local docker cache, will talk to Peter about it. Squid