Resiliency Synthesis Nov 21, 2017
ARC reviewed three resiliency proposals for Beijing OOM Already plan of record for deployments Adds HA features from Kubernetes Consul health monitoring CHAP (Common HA Platform) MUSIC: multi-site active-active failover (based on Cassandra, zookeeper) HAL – recipes (based on zookeeper) Conductor – site selection Distributed KV store Move configuration from databases to KV store Could be consul or etcd; proposal recommends consul based on tool support
Synthesis The base platform will inherit significant resiliency features from OOM When we move from HEAT template deployment in Beijing release Primarily focused on intra-site deployments Could be augmented by CHAP in multisite deployments Distributed KV store could help with multiple database problem All three could work together to improve system resiliency One potential issue: can we get to a common health check/KV/data synchronization solution? Consul, etcd, zookeeper?
Open source data synchronization services Three similar services: Consul (hashicorp) Provides service discovery, failure detection, KV store, and multisite replication Zookeeper (apache) Provides configuration information and distributed synchronization Etcd (coreos) Provides a data store across a cluster of machines Can we agree on a common tool? Do we need multiple? Do we need more information/discussion? Proposal: encourage Consul, as it’s common to two of three proposals, already part of OOM PoR, and has good tool integration
Possible new project? CHAP does not currently exist in ONAP. Is it something we would like to see proposed as a new project? Is it an enhancement to OOM?