Jump to navigation Jump to search
Thanks ops for handling it!
Proton is still shadowing Electron, right? So
- does this mean that Proton could not keep up but Electron could? That seems surprising given that they only differ in their method of remote-controlling the headless Chromium instance actually doing the rendering; plus Proton actually has a queue system for limiting the number of simultaneous renders, while AFAIK Electron doesn't.
- did this affect availability of the PDF endpoint? Again, that would be suprising since it should only be getting mirrored traffic and is on a separate server.
- in theory the queue system in Proton should be able to handle high load; do we understand where that went wrong? (Was the traffic so high that the queue manager itself got overloaded? 10 reqs/s does not seem like a big deal for a node service that only does scheduling.)