This project is designed to facilitate the standing up of basic Gluster storage-clusters. Templates can be used to create a storage-cluster of arbitrary size and layout: 2-way and 3-way mirrored-configurations laid out across both two and three availability-zones have been tested. Templates deploy the cluster nodes using AWS's AutoScalingGroup functionality. In the event of a node-failure, the AutoScalingGroup will detect the failure and build a replacement node.
It is generally assumed that this automation will only be used:
- In regions where EFS is not yet available
- In use-cases where EFS is insufficiently-performant (typically small- or medium-sized, active data-sets: see the vendor's Performance guidance-document)
- This automation has only been tested using the SPEL CentOS 7 AMIs. However, should_ also work on instances built from MarketPlace or other CentOS 7 AMIs, both SPEL and generic/MarketPlace Red Hat 7 AMIs or Amazon Linux 2 AMIs.
- While these templates leverage watchmaker to apply STIG-hardening to the deployed instances, FIPS mode is specifically disabled. Node-recovery is adversely-impacted when nodes are run in FIPS-mode.
- Automated node-replacement: Gluster's in-built security prevents replacement-nodes from either being automatically-admitted to a degraded cluster (one missing one or more members due to failure) or automatically assuming the role of a failed node.
- It is critical that the cluster be monitored for node-faults and rebuild-events — particularly if the cluster hosts critical data.
- In the event of a node-fault/rebuild-event, it will be necessary for an administrator to log into a serviving node and execute cluster recovery procedures. The documented procedures describe how to admit a rebuilt node into the degraded-cluster and use it to replace a failed node.
- Backup-automation is not currently built in (though is being considered for future iterations)
React
Typescript
Django
Laravel
Facebook
Microsoft
Google
Alibaba
D3
Tencent