Apologies if this seems easy, but I just can not get my head around the issue.
Let’s assume we have successfully deployed to Digital Ocean SSD cloud service, but now having made some code changes we need to update the server - but here is my concern how can I update the server if there are existing users on the server, I have not used Digital Ocean yet but assume it is like Heroku deployment which takes the server down, updates it and then restarts the server.
So I am wondering how do existing public Meteo apps deal with updating of code?
You start the deployment process
A bundle is made locally and sent to the server
When the server has received it it starts the build process, which stops the current server
When the build is complete the server is started again
On our own product, when there’s no big db modification to run on startup the downtime is about 5-10 seconds. The client reconnects automatically and the window refreshes if necessary (if client code has changed). All in all this is a really fast and efficient way, but there are ways to improve things:
In our case users can write comments and stuff, reloading might wipe everything they were writing, so saving the content of an input into a session variable for example and putting it in again after a reload helps. Conceptually I mean caching the state of certain elements into something that can survive a hot code push.
There’s also the possibility to use something along the lines of grounddb: https://atmospherejs.com/ground/db
When the client goes offline he can still work with ‘grounded’ databases. After he reconnects the data modifications and methods are simply resumed and sent to the server.
All in all, deploying new versions of a meteor app is a smooth experience and there are always ways to decrease the problems it might cause to users.