Hi,
Do you have suggestions for kernel tweaks for getting the most out of a RAM limited system?
I am running a service requiring 2GB of RAM (netbird) on a VPS which has just 1 GB of memory. I am doing so because I am a stingy bastard and I use only free VPSs for my personal use so I get what I am paying for.
Because of this hardware limit in about 12 hours from service start I begin swapping a bit too much. This would still be manageable but soon the hypervisor gets really pissed and steals up to 90% of the CPU. So the only solution so far is restarting the docker containers every 12 hours (not great, not terrible).
Looking to improve this, Iam now experimenting with ZRAM and swappiness and it seems some benefit can be achieved by using some of the Linux kernel feaures. Is there anything else I should look into?
Wow you’re right. You are a stingy bastard.
Never run anything on a free tier if you want to keep it. My legit Oracle account got flagged and deleted for no reason so they’re dead to me, and I’ll tell anyone who will listen. They’re thieving assholes and will snatch away free tier the moment it’s convenient for them.
Do you really need netbird? What for?
Spend some money, consider it the cost of education
To tag onto this… for a low end VPS I love Hetzner. No, they aren’t the cheapest, but the cheaper options are generally bottom-of-the-barrel unstable garbage or require promotions that expire. Hetzner is stable, professional, and cheap. Great choice for a simple personal VPS.
Netcup is cheaper (as far as I know) and really reliable too. Also based in Germany. I use one of their VPS to host my stuff.
Nice! Love to learn new options.
Can really recommend them. They guarantee the lowest price for VPSs and are very reliable. My company (more than 250 people) uses them to host some stuff too.
I’ve only had good experiences. Very reliable.
I’m going to chime in and recommend Racknerd. Inexpensive and rock solid for me.
As long as you’re not running Plex lol
I use Vultr and OVH.
I’ve heard good things about OVH.
If you’re using Oracle cloud (just guessing based off 1GB), they also offer free ARM VMs with 24GB of RAM, and netbird looks to support ARM.
Yes, this is a possibility. the ARM VPS is already running something else, but if I manage to run netbird behind a reverse proxy I can also move it there. BTW there are also 1 GB free VPS on azure (for students) and Google Cloud, but you guessed right.
Yeah good luck getting one. I tried for a month before giving up.
Turn on swap! That way more of your actual ram can be used by the application. Yes, it will probably be slow as crud but it should use more memory than it has. If swap is already on… sorry
I forgot to mention, I had plenty of swap available, now I disabled swap to force zram usage. I still need to see what happens running with both, it’s hard when each trial takes 12-24 hours to show its result.
Oh yikes yeah I would not be good at that pace 😅 I know you can rebuild the kernel with less features to reduce ram usage, but other than that I would be looking at what you’re already doing.
Have you considered removing docker from the equation and running directly on the host os?
Running without docker is out of question, is a bundle of 6 docker containers. Deployment and management without it would be too complicated. Luckily somebody in another reply made me realize that the RAM eating container (cockroach DB) is far less essential than I thought and I can look for a replacement.
Awesome, lol also you may be able to make a new email and free vps just to run cockroach db
Oracle free tier is 24GB of RAM. I hate Oracle as much as the next person, but worst thing they do is shut it off and you have to migrate.
Yes, this is a possible fallback plan.
why use docker here? you’re just adding layers of abstraction in an environment that can’t seem to really support them.
that said, switching to 32bit linux, if the VPS supports it, will save you memory.
deleted by creator
I use Docker on a VPS with 512MB of RAM
🙂💄🎨🤡
This service consist of several docker containers, without docker I would not even know where to start for deploying it. Maintainance as well would be a mess, totally not an option
I think the advice should be taken to heart here — you’re dealing with a userspace problem but you’re trying to get the kernel to make it all better.
You’ve already mentioned the two big things, compressed RAM and swap; optimizing userspace (or paying for more RAM) may be the only option at some point.
If you want to get creative, is there a reason you can’t use a local computer for some of these services? An old raspberry pi or similar could potentially run some of your services. You could run some containers on your home server and call it a day. Quick search turned up this https://www.linuxserver.io/blog/routing-docker-host-and-container-traffic-through-wireguard
Get an oracle cloud account and use their free arm server. You’ll have 24 gb of ram to play with. For free.
Thanks for the suggestion, but my 24 GB are well employed already. I wanted at least to outsource the VPN manager to a smaller VPS.
It seems to me you are already looking into all the possible options besides getting a VPS with more RAM. I am a bit confused on why you are seeing increased CPU usage by the hypervisor for this though.
The server is clearly overloaded, as soon as I start using some 10% of CPU frequently for some minutes (due to swap operations), the Hypervisor starts to throttle my instance and this of course makes the thing worse with an avalanche effect. When this happens steal time displayed from top can go literally as high as 90%.
On AWS they have something called “bursting”. Basically they will let you use 100% of your vCPU, but not all the time. If you use it constantly they start to throttle you. That’s explicitly stated when you rent an EC2 instance (which is their VPS). Perhaps your provider is doing something similar.
Yes exactly, burstable instances it’s common jargon for AWS and GCP, but applies to all major providers.
Yeah, those are great for something that will be sitting idle most of the time, awful for anything else.
Ask the provider not to throttle you then? Are you not entitled to the entire cpu?
I use only free VPSs
So probably not entitled to anything from them.
I believe it’s a shared VCPU intentionally, I will recheck the terms and conditions, but I think I am not in the position to claim much.
zram or a cloud instance with more ram https://paul.totterman.name/posts/free-clouds/
In my experience kernel tweaks aren’t going to be a major change on memory usage. Most distros are meant to be full featured and not necessarily lightweight. So unless you are already running a minimalist distro, make sure you don’t have bunch of background services running you don’t need. I can recommend using Debian Minimal iso’s, they require 256MB of mem. Depending on what features you enable you could use a lot more.
Thanks for sharing your experience, indeed the distribution is relevant here. I am running Arch (BTW) on this VPS which idles at about 300 MB with dockerd and containerd, I am not sure how does exactly compare to Debian on RAM usage (I have a couple of other VPS running debian which seem to use a little bit more RAM but it could be because those images are bastardized by the addition of cloud provider services). In any case my setup is pretty minimal, to get some large benefit there I fear I should use something without systemd :/
Yeah the 300meg isn’t going to get much less. Switching to Debian won’t change much there. Perhaps you can look into running a minimalist container distro if you are just using the machine for that. I personally want to check out Talos, there’s also RKE and Burmilla. No experience with them, to me the memory doesn’t matter much because I run a homelab. So I currently just run Debian and k3s. On my systems the containers are actually what gobbles up all the memory. If you’re using public container images, there’s a good chance the memory configuration on them isn’t optimal. Especially JVM services are a lot of the time configured to just use whatever is available. If you give them less memory they will do more garbage collection. So if CPU is less an issue then mem, that could be worth looking into (it’s just parameters you can pass on startup). Hopefully any of this is of use. Good luck :)
Thanks again, I will look into your suggestions, never heard of neither Talos, RKE or Burmila. Indeed I should also look if I can do in the containers. The problem is only this database “CockroachDB” which is extremely memory hungry, maybe I can change something there.
CockroachDB recommends at least 4GB of RAM according to their documentation, I don’t think you can shave that down to a nornal level.
However, the CockroachDB dependency doesn’t seem to come from Netbird itself but from Zitadel. You can try using another identity provider listed here if you can find one that takes less RAM, or even use external services to handle auth for you. I can already tell you that Keycloak is quite resource hungry at times, but the entire thing seems to just do basic OAuth2 or OIDC authentication so there are plenty of lightweight alternatives.
I’m not sure what your use case for Netbird is, but if you just want a peer to peer network and don’t care for the complex IdP integrations, Tailscale provides a similar peer to peer network and is much lighter on resources. You can use headscale to run the server and then use the official Tailscale clients to connect to it (the default setup will use Tailscale’s servers at least a little, but you can host everything yourself).
Thanks, this is a really good point, I can try to replace the identity provider! I did not realized that cockroachDB was only a Zitadel requirement! There are many great alternatives for mesh VPNs, netmaker, nebula, and headscale as you mentioned and all of them are much lighter. I ended up hosting netbird as it is natively able to traverse my corporate NAT (maybe headscale could do it as well, I did not try it since I do not like having to configure registry keys on windows clients and losing the kernel wireguard speed on linux clients) .
CockroachDB is a clustered version of PostgreSQL you probably should be able to replace it with that. But running a full RDBMS with the resources you gave is not great. SQLite would be a better fit for the resources available if the tools you run support it.
One of the answers in this discussion made me realize that this database is required only by the identity provider which I can change from the default. Considering this is a far less essential dependency than I thought I will get rid of it completely.
Sounds like an excellent suggestion.
If you’re idling at 300mb with containerd running, you’re not getting better than that with a modern general-purpose distro. As others have said, switching to another vps’ free tier that offers more is your single best bet by a mile. About the only options you have on this one are compressed ramdisks being used as swapfile (zram) and literal disk swapfile to get you the rest. It’ll be very slow though if you have to load half your workload on shared platter swapfile.
Try using zram? You could probably get around 5gb of ram if you compress it. https://wiki.archlinux.org/title/Zram
I am experimenting with ZRAM, it is indeed better than ZSWAP, that’s why i am asking if any other kernel features can help.
Zram (compression), uksmd (deduplication) and swap.
For a different project, but in my blog I document how to set those up: https://moonpiedumplings.github.io/projects/setting-up-kasm/
What are you trying to do? I read “hypervisor” and am not sure if you mean the one managing your VPS or if you installed one in your VPS.
The one managing my VPS, controlled solely by Oracle corporation
