Back to Blog

How Mesh brought seven multi-region relays across six servers into one Dokploy dashboard

Will

November 18, 2026 • 5 min read

How Mesh brought seven multi-region relays across six servers into one Dokploy dashboard

Mesh Network Association runs a decentralized, end-to-end encrypted messaging network supported by independent relays.

With relays running on Azure VMs (recently expanded to five global regions: Canada Central, US West, Sweden Central, East Asia / Hong Kong, and South Africa North) alongside two on its main server, Mesh uses Dokploy to bring deployments into a repeatable workflow without changing how its relays communicate.

IndustryLocationKey benefit
Nonprofit – Decentralized communicationsFranceManage independent servers from one dashboard

Key stats

7+ relays6 servers0 downtime / 0 failed deploys
Two on the main host and five distributed across global Azure regionsOne main server and five independent Azure remote serversBrought under Dokploy while preserving their identities and eliminating configuration drift

The challenge: More servers, more moving parts

Each remote Mesh relay has its own public address, identity keys, and persistent databases. Each needs the correct TCP and UDP ports and a deployment configuration suited to its host. After all, you can't replace a relay like a stateless web application.

Mesh already operated five relays on separate Azure virtual machines, alongside two on its main server. The association wanted a clearer way to manage services across those machines while keeping existing identities and data intact. 

They needed to move to a new deployment workflow without having to start the Azure relays from scratch.

The public endpoints, keys, and databases already belonged to functioning relays. Mesh had to retain that state and make sure it was easy for maintainers to identify which server would receive each deployment – especially with multiple independent machines running the same relay software.

Prior to Dokploy, relay deployments were carried out through separate, manual SSH sessions and disparate shell scripts across each virtual machine. This manual approach created three critical friction points: configuration divergence between servers, the tedious operational burden of logging into each machine sequentially whenever a new relay image was released, and – most critically – the risk of identity loss.

Because each Mesh node relies on persistent cryptographic identity keys mounted in Docker volumes, any manual error or accidental volume overwrite during redeployment would destroy the relay’s established cryptographic identity on the network.

“Our biggest challenge was operating isolated relay nodes across multiple continents without introducing complex cluster dependencies or wasting limited non-profit compute resources. Dokploy gave us the best of both worlds: each relay remains fully independent and isolated for resilience, while our lean team of two infrastructure volunteers manages, inspects, and upgrades the entire fleet through a single, elegant interface.”

– Marc Moreau, President, Mesh Network Association

The Dokploy features Mesh uses

Remote ServersDocker ComposeDeployments, Logs and API
Select and manage independent Linux servers over SSH.Define each relay’s image, ports, restart policy, and persistent volume.Review manual relay deployments, inspect output, and trigger website deployments via GitHub Actions.

How Mesh uses Dokploy to manage its relays

Mesh organized its installation into one Dokploy project and one production environment, with separate Compose services for the website, local relays, and each Azure relay.

A service shows its destination server, so the team can deploy the website without touching a relay and update one remote machine independently of the others.

Mesh manages its website and independent relay services from one Dokploy production environment.

Each relay’s Compose configuration records its container image, advertised hostname, transport ports, restart policy, and persistent volume.

These settings stay explicit for each host. Even when two independent VMs use the same volume name, those volumes stay as separate stores, preserving the different identities of the relays that run on them.

Each relay uses its own Docker Compose configuration, with persistent volumes preserving its identity and data.

To bring Azure into the workflow, Mesh registered each VM as a remote server using dedicated SSH keys, verified the connection, and ran the server setup.

Dokploy was the interface where they selected a host, reviewed a saved Compose configuration, deployed, and checked logs. The hosts remain independent; they are not part of a shared Docker Swarm cluster.

The team had to be careful during the migration. They inventoried the existing containers and volumes, backed up keys and databases, and then stopped the original containers before deploying replacements through Dokploy.

Mesh reused the existing images and external volumes, ensuring only one container accessed a relay’s data at a time. Afterwards, the team verified identity responses and TCP/UDP connectivity. Both Azure relays retained their identities.

The team completed the migration of existing production relays in less than a day with zero network downtime.

Since the transition, Dokploy has eliminated dozens of repetitive manual steps: provisioning a new relay node has dropped from ~45 minutes of manual server setup down to under 5 minutes, while rolling out network-wide relay updates takes less than 2 minutes instead of 30 minutes of manual SSH logins.

Crucially, failed deployments due to configuration typos or environment mismatches have dropped to zero.

A different approach to multi-server deployment

Unlike applications that replicate identical services across a cluster, Mesh needs independently identified relays hosted on separate machines.

Dokploy manages their lifecycle over SSH, while the messaging traffic travels directly between the relays. If the management dashboard becomes temporarily unavailable, the running relay containers can keep operating.

This architectural choice was deliberate. In a decentralized, censorship-resistant messaging network, nodes must be strictly fault-isolated. Shared clusters (such as Kubernetes or Docker Swarm) introduce cross-regional network latency penalties, cluster quorum fragility during intercontinental network partitions, and heavy resource overhead on lightweight VPS instances.

Dokploy’s independent Remote Servers model ensures that if one cloud provider or geographical zone experiences an outage, no cluster consensus is broken and the remaining relays continue operating without disruption.

Dokploy's Remote Servers interface lets Mesh manage geographically distributed servers independently.

Mesh also combines automated and deliberate deployment workflows. Its website is deployed through GitHub Actions using Dokploy’s API after a new container image is published. Relay deployments remain manual, letting the team review the image and configuration before changing a service carrying network traffic.

What’s next for Mesh?

Mesh is preparing to expand its relay network across more hosting providers. With Dokploy, the association can reuse its deployment process:

  1. Prepare a Linux server
  2. Set its destination and public endpoint
  3. Attach its own persistent storage
  4. Deploy.

The team still needs to pay attention to capacity planning, backups, and release checks as the network grows, but the operational workflow is already in place.

Mesh now has one place to manage a growing set of independent relays, without losing the relays’ identities or the independence of their servers, which leaves more time to develop the messenger, test its protocol, and support the people who operate the network.

Today, the infrastructure is maintained by just two core volunteers. Having already expanded from two to seven active relays across Microsoft Azure (Canada, US West, Sweden, Hong Kong, and South Africa) and self-hosted hardware, the association plans to open relay hosting to community operators and additional cloud providers in 2027, further expanding global geographical and legal diversity.