At the minute, we're using one server to handle all of the traffic to the platform; this isn't too bad because it has quite high specs and has done a good job at handling the traffic. To better support expansion in the future, however, it should be switched to a distributed method across different serves which handle different tasks. For example, dedicated database servers, and dedicated MediaWiki servers. Since we're caching via Ezoic, there doesn't need to be a cache server.
Description
Description
Status | Assigned | Task | ||
---|---|---|---|---|
Resolved | OriginalAuthority | T26 Distribute traffic over different servers | ||
Resolved | OriginalAuthority | T27 Switch file storage to Amazon S3 | ||
Resolved | OriginalAuthority | T28 Launch a MW2 Server | ||
Resolved | OriginalAuthority | T41 Move database to its own server | ||
Invalid | None | T29 Set up and configure Graylog | ||
Resolved | OriginalAuthority | T46 Launch a HAProxy Server |
Event Timeline
Comment Actions
Update on this, we're not caching through Ezoic anymore, but we could still benefit from this; we still don't need a cache server, because we're caching through Fastly, but we can probably use HAProxy as a load balancer—it offers better configuration than Hetzner.
Comment Actions
This will most likely be done by virtualisation with Proxmox, running on a bare metal server.
Potential configurations:
Server | RAM | CPU | Storage |
mw101 | 12GB | 3 | 150GB |
mw102 | 12GB | 3 | 150GB |
db101 | 16GB | 2 | 100GB |
task101 | 4GB | 1 | Shared allocation with mw101 & 102 |
puppet101 | 4GB | 1 | 50GB |
mem101 | 4GB | 1 | 10GB |
This would represent a small overallocation of the resources available, but this shouldn't really be an issue given that not everything will be hitting heavy computation power at the same time.
Comment Actions
The server is now set up using Proxmox, configuration is ongoing, and it is expected by EOW that we can have this live.