Montiapm is Kadira but with a few added features like production source maps. It’s being offered at a great price now so you should definitely use it.
Agree here, Terraform rocks and is key to our decision to move to self hosting. I’m not even talking about how easy it is to do a small change and then let it figure out what exactly needs to be changed vs otherwise re-running your whole process again.
It’s crucial for our success and I can’t give enough praise to what the guys at HashiCorp have build. For more check here: https://www.terraform.io/intro/index.html
Disclaimer: not affiliated with any of the companies names, just a happy user
Have heard about Terraform but have not as yet got my head round it yet. A tad daunting the one time I tried reading through. Need to just get stuck in…
You don’t happen to have any writeups or scripts regarding how to go about this? I’ve been looking into Terraform as well, but as @hemalr87 says, it is quite a daunting task it seems.
Terraform is awesome! Qualia uses it for everything
This thread might be of interest:
For those interested in Terraform there is indeed a ton of excellent stuff out there in the form of blog articles.
Just google it and go step-by-step
Up to now it has not been a lack of features that have annoyed me, but after getting my account suspended with absolutely no alerting, warning or follow up within 19 hours from support, I have been looking at alternatives and there is no place in hell I’ll opt for a service that has a poorer response time than both linode and vultr where the cost is a fraction of the Galaxy containers.
I don’t think Galaxy lacks features, but the slowness (see https://www.erichartzog.com/blog/aws-vs-galaxy-for-meteor-hosting ) is really bad… can anyone explain why exactly? And is it going to improve? (/cc @hwillson ?)
I have several deploys on Galaxy (0.5 ECU, $40), and one on a DO droplet (2v CPU, $40), the DO one is 10x faster somehow… (and it feels like increasing the .5 to 4.2 doesn’t improve the speed that much…?)
Are there any tools/scripts to do some speed tests?
We switched one of our apps from Galaxy to Waves which uses ELB . The results were pretty amazing like mentioned in that blog you reference @batist. The clients were absolutely delighted. There were a few kinks to work out to get everything working smoothly but hopefully Waves can incorporate those so new users don’t have any issues.
Other than performance we also got
- TFA (both in waves & AWS)
- Alerts (via cloudwatch)
- When you deploy the immutable deploy gives even less of a refresh than Galaxy.
It’s not quite as convenient and easy as Galaxy but the extra performance and features are well worth a bit of devops.
We’re in the process of switching our remaining project.
Would you mind sharing a little more about those difficulties?
My setup is a DO instance with MUP. I’m considering switching to Waves because of autoScaling and rolling updates.
Did you manage to build a CI pipeline?
We change our settings with each release as that’s where we store our version number. In Waves if you change your settings your machine will restart. We created an EB extensions which reads the settings from our settings file when the machine starts and sets the environment with this. Otherwise you’d have to find another way of setting the version.
If you want to know about the memory usage of your instances it’s a bit of a pain, although this is down to how EC2 works. By default there is no metric for measuring RAM so you have to install a package that then reports it to cloudwatch. Again, this is done via the EB extensions. If you need more details we can post our config here. It was a royal PITA.
To create a dashboard in cloudwatch that updates itself after a new deploy requires a little lambda script to poll every so often to see if the metrics running are still valid, and if not try and update them. This is also a bit annoying and I’ve not got round to actually finishing this, so hopefully Waves can figure out a way to automate it.
Also, if you use cloudflare, you’ll need to find the right moment to reset the cache. What we did was poll our server from our CI script (we use codeship) after deploy. As soon as the version we just deployed is responding we then ask cloudflare to clear it.
Hope that helps
Using Google Cloud , so far no issues , using load balancing , with haproxy routing to instances on different zones . i build the apps using the meteor build , and publish tar file to google storage and after that , have scripts on the app servers to just pull the build and deploy it . App runs using pm2. Rotate the logs.
Essentially the set up is as follows
Google Load Balancer ( ssl enabled) -> haproxy -> app deployed on vms in multiple zones -> pointing to a mongo db + Redis server + google storage for image uploads . Everything behind the load balancer have internal IP addresses.
This set up is very simple and i can migrate from one host to other without any problems …
Been wanting to try out Waves for a while but it requiring admin access to AWS has put me off so far.
I wish they would create a minimum permission needed IAM user requirements guide. I guess I could create a separate AWS account for just this purpose… I’ll ask them and see what they say.
Also wondering about static IP’s…
You can add specific permissions instead of admin:
The tutorial is now updated:
Hey @marklynch, would meteor-apm be a solution for RAM monitoring ?
Also, did you come up with a solution for cloudfare ?
I am using cloudfront but it is the same problem.
Hey, apologies @sabativi, was away from the forums for a while. Meteor-apm is specifically for using the mdg fork of kadira, so requires using galaxy. We use the monti-apm fork. But yes, it reports RAM usage. We wanted to get an overall dashboard running so hence using cloudwatch.
What problem with cloudflare are you referring to ? We use it and don’t have any issues.
Hey, thanks for your response,
Was a bit in a blur when sending this message, but now it is clearer, cloundfront works like a charm
Thanks for your response
We use Galaxy. Yes, more expensive until you factor in the cost savings of a reliable provider that works. We run a hack for autoscaling. I hate it. That is the pain point that would push me to leave.