How can your CloudStack users build on the solid, stable CloudStack deployment that you built? The presentation will examine / demonstrate emerging open source technologies such as PaaS (CloudFoundry for example), Docker, CoreOS, Apache Mesos and Kubernetes and how they can run happily and efficiently on CloudStack
A followup to the CCC13 presentation "How to Run from a Zombie" (https://www.youtube.com/watch?v=nFaRl5hgsG0) which explained that the CloudStack architecture violates the CAP theorem by attempting to be simultaneously consistent and available. These conflicting goals cause the system to misallocate resources, orphan device configuration elements, and nondeterministic application of infrastructure configuration.
This presentation will introduce Hoke, a resource management library designed to address these issues by separating consistent automation control and availability-oriented orchestration operations. It provides a lock-free, consistent automation control plane using state-based replication with an orchestration tier. Finally, it distributes itself across the partitions of a CloudStack cluster to scale out workload and isolate the impact of infrastructure failures.
CloudStack's storage infrastructure does not currently allow for multiple VMs to attach to the same data disk simultaneously.
Interoute, however, has a critical use case whereby there is a necessity to allow multiple VMs to perform such an action.
In this situation, we were constrained to using an existing version of CloudStack (4.5), so no code changes to the "core" software could be made.
We decided on an approach that leveraged CloudStack's extensive plug-in framework.
Using CloudStack's ability to support plug-ins, we developed three plug-ins to solve the problem:
1) A plug-in that extended CloudStack's API to support additional functionality.
2) A GUI plug-in for root admins to set up storage and VLANs.
3) A second GUI plug-in (typically used for standard users) to create volumes that can be directly accessed via the OSes running in the applicable VMs.
Talk is about how organisations with an existing directory service such as Microsoft AD or LDAP can use CloudStack by using LDAP, Single Sign On using SAML and writing a custom authentication plugin to integrate CloudStack in their custom environment for example by a 2-factor authentication system based etc.
For resource allocation, currently CloudStack uses heuristic/threshold based algorithms which are error prone and may not fit best for all environments. In addition to this different kind of applications have different kind of resources (compute, storage etc.) requirements which is tricky and hard to allocate. With some initial learning provided, Reinforcement Learning based approaches will learn the best policies for system automatically. If there are some changes in environment than it will automatically detect changes and will try to modify policies which are more suited for changed environment. Reinforcement Learning based techniques can also be utilised for VM consolidation in CloudStack. VM consolidation helps in operational savings of data center.
CloudStack performs well with default configuration for most of the deployment. But every system has a limit, beyond which the performance degrades. To get rid of these performance issues we have to tune the management server by tweaking some configurations. In this presentation, Sudhansu will discuss various cloudstack configurations and the load that will make management server exhausted. We will also discuss common issues and solutions. Ex: A cloudstack Simulator based environment with 4K hosts, 4K accounts, 12K Vms, 8K router Vms, 2 management server nodes, 8G heap size started showing hosts in alert and disconnected state after 20 minutes of operation. The root cause was found to be less number of database connections.The configuration that determines the maximum active connection is 'db.cloud.maxActive'. To address this issue the we configured this to 1000 (default 250).
The aim of this session is to introduce people to CloudStack development process, tools etc and show how they can start contributing to CloudStack as developers. The session will also feature development of small feature(s) and a plugin.
In this presentation Remi Bergsma shares his tool box of scripts that come in handy when you're working with CloudStack on a day-to-day basis. The scripts were written to automate operating the Schuberg Philis Mission Critical Cloud, which is built on top of CloudStack. They're now available for anyone to use.
Getting overview of what runs where, automated migrations, handling an outage, rebooting a cluster without impact (patching!), massively upgrading system VMs are just some of the automated tasks that Remi will talk about.
A must-see if you're operating a CloudStack cloud.
Remi Bergsma (NL) is an Engineer that loves building and managing infrastructure based on Open Source technologies. His goal is to automate what can be automated, so everything works rock solid and he has the time to explore new tools and techniques (that usually result in even better... Read More →
Friday October 9, 2015 9:30am - 10:00am IST
Liffey Hall 2
OSPF is a routing protocol for Internet Protocol networks. It uses a link state routing algorithm and falls into the group of interior routing protocols, operating within a single autonomous system. Compared with RIP, OSPF can provide scalable network support and faster convergence times. OSPF is widely used in large networks such as ISP backbones and enterprise networks. Considering that in large organizations where the end-to-end access is a requirement, the current VPC model brings more complexity to management and hands-on configuration beside the overhead involved in the NAT address translations processing. This presentation will bring some possibles approaches to enable dynamic routing using the OSPF protocol for IPv4 and IPv6 and simplify the network administrator life safely and quickly.
Cloudstack has been changing and sometimes it is frustrating to get the most current information about the ways things work in cloudstack. This tutorial is about putting together a API plugin and a UI plugin. Will also delve into plugin support provided by cloudstack. Walking thru the intricacies, troubleshooting (various errors and what will they mean you missed).
Currently Secondary storage in CloudStack plays key role in areas like snapshot, template management and also to some extent volume management too.But there exists well known performance limitations, especially in huge clouds with hosts spread across multiple pods, associated with secondary storage available across entire zone.Users often complain about slower snapshots, cold migration or deploy operations due to performance limitations of secondary storage which is not optional today per design without flexibility.This paper proposes an alternative design for the problem, by making secondary storage optional component in CloudStack deployment.Goal is to leverage existing primary storage in clusters across the zone to host snapshots, templates etc. and improve speed of all associated operations.Will present all limitations posed by this design change & solutions to address limitations.
CloudStack provides software engineers with the ability to extend many aspects of its behavior. Come learn how to develop a CloudStack API plug-in to enable your business to integrate third-party products more deeply into CloudStack.
Virtual Router(VR) is a key component in the functioning of CloudStack. VR also tends to become the pain point for many users, more so during upgrades. In this presentation, Kishan Kavala will propose and discuss options to improve VR operations and ways to make upgrades smooth with minimal disruption.
Making Cloudstack management servers total redundant with the use of MariaDB Galera Clustering and Load balancing technologies. Glenn will go through the technologies used, the best way to configure them, effectively getting the HA benefits of Galera clustering auto-failover, full consistency between DB nodes and no DB slave lag.
We know how to import the network piece of cloustack VR to the network. If the VR goes down for a moment then there is downtime for the cloud. Most of the cases VR is failed to start or go down in upgrading the cloustack to new version due to changes happened in cloudstack code changes or script/system.iso changes. In post upgrade if the VR start is failed then this creates tension to cloud operator.
This paper discuss about the how to debug the VR to bring it up and running when it is failed to start or correcting the VR services which caused the VR service failures. It also gives the few productions issues examples where the VR brought up successfully with in short time.
