# durable-objects
- g
Greg Brimble | Cloudflare Pages
04/04/2021, 2:32 PMCould and probably should be sanitizing on the entrypoint Worker script. - g
Greg Brimble | Cloudflare Pages
04/04/2021, 2:33 PMI see my network of DOs as my trusted VPC essentially. Anything in there, I trust. Anything incoming to the Worker Script, I don't. So like @User said, the Worker Script is my gatekeeper. - v
vans163
04/04/2021, 2:34 PMOh another cool feature would be a way to keep a DO alive 24/7 - v
vans163
04/04/2021, 2:34 PMTo prevent wierd edge cases - g
Greg Brimble | Cloudflare Pages
04/04/2021, 2:35 PMGot an example? - v
vans163
04/04/2021, 2:35 PMTho nm that's not a solution - v
vans163
04/04/2021, 2:35 PMEventually some act of god or system maintenance will bring it down - e
eidam | SuperSaaS
04/04/2021, 2:35 PMIts serverless, it should be brought down 🙂 - g
Greg Brimble | Cloudflare Pages
04/04/2021, 2:35 PMYeah, you should probably be just re-getting the stub every time you can break syncronicity. - v
vans163
04/04/2021, 2:36 PMWhen using KV worker storage, instead of the DO storage. And fetching out a copy of the KV data - g
Greg Brimble | Cloudflare Pages
04/04/2021, 2:36 PMV. good point lol - v
vans163
04/04/2021, 2:36 PMIf the DO dies and ups again. Itl read its initial state from global KV but global KV might be behind - v
vans163
04/04/2021, 2:36 PMDuring these seconds - g
Greg Brimble | Cloudflare Pages
04/04/2021, 2:36 PMAt that point, you've essentially got an EC2 instance haha - v
vans163
04/04/2021, 2:36 PMBut I dont wanna deal with amazons crap lol and overprovisioning - v
vans163
04/04/2021, 2:37 PMAnd bandwith charges that are outright fraud - g
Greg Brimble | Cloudflare Pages
04/04/2021, 2:37 PMIf you can't deal with your data being wrong, you have to use DO (or some other provider). KV just won't work for it. - v
vans163
04/04/2021, 2:38 PMI can deal with data being wrong but a few users might get stuck jobs - v
vans163
04/04/2021, 2:40 PMhttps://zod.tv, basically this is what we doing. We want to put job queues for video transcoding jobs on cloudflare. Before using a simple POC with workers unbound the problem was our workers polling for work. 1 worker ate up a month of CFW requests lol cuz it polls every 3 seconds - v
vans163
04/04/2021, 2:41 PMIt needs to be as responsive and realtime as possible - v
vans163
04/04/2021, 2:42 PMWe plan to have atleast 1000 and scale to 100,000 zodworkers later - v
vans163
04/04/2021, 2:42 PMThe workers will keep a WS open 24/7 - v
vans163
04/04/2021, 2:42 PMSo the walltime billing is icky - v
vans163
04/04/2021, 2:43 PMItl be using 0 cpu just ram to keep a socket open and some basic kernel tcp time - k
Kat
04/04/2021, 2:43 PMSounds like it would be drastically cheaper to just run an EC2 instance lol (even though you don't want to) 😛 - v
vans163
04/04/2021, 2:43 PMDont want the service going down - v
vans163
04/04/2021, 2:43 PMEver - v
vans163
04/04/2021, 2:43 PMEc2 is unreliable - v
vans163
04/04/2021, 2:44 PMRunning 10 ec2 instances across 3 cloud providers - k
Kat
04/04/2021, 2:44 PMWhat about like Kubernetes or something? 🤔