Easy Geo-Redundant Handover + Failover with MARS + systemd

on YouTube ↗23m

" summary

Long-distance, transaction-logged replication using MARS transaction-logged replication and a new systemd unit-template generator automates geo-handover so a secondary can become primary while preserving strict local consistency and propagating sibling-tree state. Production MARS backs thousands of LXC containers and several petabytes at 1&1 IONOS since 2014; the new systemd interface is functional but not yet deployed at scale and currently uses breakable locks to work around a reentrancy deadlock in systemctl. Metadata scalability tops out around 14 generated resources today.

AI-written summary · about

" chapters (8)

1. 0:00Start
2. 0:21Easy Geo-Redundant Handover+Fallover: Agenda
3. 3:04Cluster Management for 50 km
4. 6:13systend Unit Example Template
5. 8:30Current status of systemd interface
6. 9:25Usage of systend unit templates
7. 12:25Current status of system interface
8. 14:40MARS Future Plans

" related talks

• 01h 08mMARS: Replicating Petabytes over Long DistancesThomas Schöbel-Theuer · FrOSCon 2015
• 43mHot Deploying Low-Latency Services for 24/7 OperationMartin Thompson · YOW! Melbourne 2022
• 59mRunning a real world, mission-critical system on AzureLoek Duys · NDC London 2023
• 25mEvolution of Geo-replication in GlusterHari Gowtham Gopal · DevConf.CZ 2020
• 53mMulti master replication for painless fault toleranceStas Kelvich · PGConf Silicon Valley 2016
• 42mUsing Mesos to Drive Devops Adoption at ScaleVivek Juneja · MesosCon Asia 2016
• 47mDeploying Like a Boss: From 2 Servers to 20,000Oliver Nicholas · GOTO Chicago 2015
• 18mLeveraging Multiple Regions to Improve Site Reliability:Andrew Duch · SREcon18 Americas