Good news!
We’ve been improving our infrastructure quality and IP availability. What’s new:
1. We’re doubling down on internal IP monitoring to ensure that any IPs assigned to servers are fully functional.
2. There’s now a full new IP range on Hetzner.
Recent updates have reduced IP-related issues significantly, and we are working closely with our infrastructure partners to further improve reliability and connectivity across all locations.
Next up is ProVM. We're working on more improvements there for consistent performance across all regions (PS: try ProVM Turbo, our new high-end range of VMs).
Read in full: https://www.doprax.com/blog/how-were-improving-ip-quality-across-doprax
Do you have improvement suggestions? Let your voice be heard: https://docs.google.com/forms/d/e/1FAIpQLScSlPQl7Agr-NsnoUP9UqQefi9IN7jxxl_u1jro3LbVy9guEQ/viewform
We’ve been improving our infrastructure quality and IP availability. What’s new:
1. We’re doubling down on internal IP monitoring to ensure that any IPs assigned to servers are fully functional.
2. There’s now a full new IP range on Hetzner.
Recent updates have reduced IP-related issues significantly, and we are working closely with our infrastructure partners to further improve reliability and connectivity across all locations.
Next up is ProVM. We're working on more improvements there for consistent performance across all regions (PS: try ProVM Turbo, our new high-end range of VMs).
Read in full: https://www.doprax.com/blog/how-were-improving-ip-quality-across-doprax
Do you have improvement suggestions? Let your voice be heard: https://docs.google.com/forms/d/e/1FAIpQLScSlPQl7Agr-NsnoUP9UqQefi9IN7jxxl_u1jro3LbVy9guEQ/viewform
Claude Code & Your VM: When It’s Especially Useful
Some server tasks are faster to do by hand. Others are easier if you let Claude Code do them for you.
IIf you run WireGuard, 3x-ui, or Marzban, you'll hit the second case more often than you think:
→ Two inbounds look the same. One works, one doesn't. Instead of checking both by hand, ask Claude Code to compare the configs and the logs.
→ Some WireGuard peers haven't connected in a week, and you didn't notice.
→ Rotating a cert means changing three files and restarting things in the right order. Easy to make a mistake late at night.
This doesn't mean you stop knowing your own setup. Claude Code still shows you every command before it runs. It just saves you from checking three log files by hand, one by one.
See more examples here: https://www.doprax.com/tutorial/claude-code-your-vm-practical-use-cases
Some server tasks are faster to do by hand. Others are easier if you let Claude Code do them for you.
IIf you run WireGuard, 3x-ui, or Marzban, you'll hit the second case more often than you think:
→ Two inbounds look the same. One works, one doesn't. Instead of checking both by hand, ask Claude Code to compare the configs and the logs.
→ Some WireGuard peers haven't connected in a week, and you didn't notice.
→ Rotating a cert means changing three files and restarting things in the right order. Easy to make a mistake late at night.
This doesn't mean you stop knowing your own setup. Claude Code still shows you every command before it runs. It just saves you from checking three log files by hand, one by one.
See more examples here: https://www.doprax.com/tutorial/claude-code-your-vm-practical-use-cases
Running your own V2Ray server is only half the setup. The other half is getting your phone or laptop to actually use it.
If you're on iPhone or Mac, V2Box is a clean way to do that. Scan a QR code from your panel, tap connect, done.
→ Works with VMess, VLESS, and Trojan
→ Supports Reality and uTLS on the client side
→ Merges Wi-Fi and cellular for a more stable connection
→ Can share the tunnel locally over SOCKS5/HTTP for other devices
Full walkthrough here: https://www.doprax.com/tutorial/connect-to-your-provm-v2ray-server-from-iphone-or-mac-with-v2box
If you're on iPhone or Mac, V2Box is a clean way to do that. Scan a QR code from your panel, tap connect, done.
→ Works with VMess, VLESS, and Trojan
→ Supports Reality and uTLS on the client side
→ Merges Wi-Fi and cellular for a more stable connection
→ Can share the tunnel locally over SOCKS5/HTTP for other devices
Full walkthrough here: https://www.doprax.com/tutorial/connect-to-your-provm-v2ray-server-from-iphone-or-mac-with-v2box
Ubuntu, Debian, or Fedora for your next server? The answer depends on what you're optimizing for.
→ Ubuntu Server: widest docs and compatibility, 5 years of support, the safe default
→ Debian Server: leanest base, no Snap, ideal for small VMs running one or two services
→ Fedora Server: newest packages, but a ~13 month clock before you need to upgrade
→ AlmaLinux / Rocky Linux: RHEL-compatible, ~10 year lifecycle, no subscription needed
→ CentOS Stream: a rolling preview of RHEL, built for testing, not for a box you forget about
Full breakdown with a quick decision table: https://www.doprax.com/blog/ubuntu-server-vs-debian-server-vs-fedora-server
→ Ubuntu Server: widest docs and compatibility, 5 years of support, the safe default
→ Debian Server: leanest base, no Snap, ideal for small VMs running one or two services
→ Fedora Server: newest packages, but a ~13 month clock before you need to upgrade
→ AlmaLinux / Rocky Linux: RHEL-compatible, ~10 year lifecycle, no subscription needed
→ CentOS Stream: a rolling preview of RHEL, built for testing, not for a box you forget about
Full breakdown with a quick decision table: https://www.doprax.com/blog/ubuntu-server-vs-debian-server-vs-fedora-server
Cloudflare launched a new product yesterday. It's called Cloudflare Wallets, and it's built for AI agents (not people).
The idea: AI agents need a way to pay for things on their own. An API call, a piece of data, access to gated content. Right now an agent hits a login page built for humans and just stops. Cloudflare wants to fix that with wallets agents can use directly.
Two wallet types exist. Account Wallets are for people and organizations. Virtual Wallets are for agents, and they run through API keys. Account owners set the limits: spending caps, an allow list of merchants, a max transaction size. The agent spends inside those limits without asking for approval every time.
Right now, only handle reservation is live. You can claim a human readable handle at cloudflare.pay. Funding a wallet and letting an agent actually spend from it comes later, no fixed date yet.
Pretty cool, if you ask us.
Source: https://blog.cloudflare.com/wallets/
The idea: AI agents need a way to pay for things on their own. An API call, a piece of data, access to gated content. Right now an agent hits a login page built for humans and just stops. Cloudflare wants to fix that with wallets agents can use directly.
Two wallet types exist. Account Wallets are for people and organizations. Virtual Wallets are for agents, and they run through API keys. Account owners set the limits: spending caps, an allow list of merchants, a max transaction size. The agent spends inside those limits without asking for approval every time.
Right now, only handle reservation is live. You can claim a human readable handle at cloudflare.pay. Funding a wallet and letting an agent actually spend from it comes later, no fixed date yet.
Pretty cool, if you ask us.
Source: https://blog.cloudflare.com/wallets/
You can have AI for free.
There’s this Github repo which lists all the free LLM APIs (for example free tiers of models like Gemini). If you ever wanted to experiment with agents or automated workflows, you don’t have to spend anything on that.
Read more in our blog: https://www.doprax.com/blog/the-free-tier-llm-api-is-the-cheapest-way-to-run-an-agent
There’s this Github repo which lists all the free LLM APIs (for example free tiers of models like Gemini). If you ever wanted to experiment with agents or automated workflows, you don’t have to spend anything on that.
Read more in our blog: https://www.doprax.com/blog/the-free-tier-llm-api-is-the-cheapest-way-to-run-an-agent
The moment you start a new server, bots can (and will) discover and scan it. Most of the attacks are lazy: the bots are looking for password logins and default admin accounts.
Do these two things to protect yourself:
→ SSH keys instead of passwords. A bot can guess a password. It cannot brute-force a key.
→ A normal user with sudo instead of root for everything. One compromised session shouldn't hand over the whole machine.
Keep your system up to date to prevent bugs from being exploited, run a firewall, and let fail2ban ban the repeat knockers.
Full 6-step walkthrough with every command: https://www.doprax.com/tutorial/how-to-secure-a-new-linux-server-in-6-steps
Do these two things to protect yourself:
→ SSH keys instead of passwords. A bot can guess a password. It cannot brute-force a key.
→ A normal user with sudo instead of root for everything. One compromised session shouldn't hand over the whole machine.
Keep your system up to date to prevent bugs from being exploited, run a firewall, and let fail2ban ban the repeat knockers.
Full 6-step walkthrough with every command: https://www.doprax.com/tutorial/how-to-secure-a-new-linux-server-in-6-steps
5 Commands to Check a New VPS
You just got SSH access to a new server. Before you install anything, spend two minutes confirming you got the configuration you need:
1. Check the OS image you actually got (cat /etc/os-release)
2. Confirm the specs (nproc, free -h; df -h)
3. Look at your IP addresses (ip -4 addr show; ip -6 addr show)
4. See what is already listening (ss -tulpn)
5. Test disk write speed (dd if=/dev/zero of=/tmp/test bs=1M count=1024 oflag=dsync; rm /tmp/test)
Full tutorial here: https://www.doprax.com/tutorial/5-commands-to-check-a-new-vps
You just got SSH access to a new server. Before you install anything, spend two minutes confirming you got the configuration you need:
1. Check the OS image you actually got (cat /etc/os-release)
2. Confirm the specs (nproc, free -h; df -h)
3. Look at your IP addresses (ip -4 addr show; ip -6 addr show)
4. See what is already listening (ss -tulpn)
5. Test disk write speed (dd if=/dev/zero of=/tmp/test bs=1M count=1024 oflag=dsync; rm /tmp/test)
Full tutorial here: https://www.doprax.com/tutorial/5-commands-to-check-a-new-vps
Shared vCPU vs Dedicated vCPU
Two VPS plans can both say 4 vCPU and cost very different amounts. One word (shared/dedicated) on the pricing page explains why.
Shared: Your vCPUs sit on physical cores that other customers also use. When the host is quiet you get close to a full core. When several machines get busy at once, your process waits. Cheap, and fine for most workloads.
Dedicated: The cores are reserved for your VM only. Same clock speed as a shared core, so it is not faster. It stays the same speed when the host gets busy. Usually costs two to three times more.
Full post with the sysbench test and which workloads need which plan: https://www.doprax.com/tutorial/shared-vcpu-vs-dedicated-vcpu-explained
Two VPS plans can both say 4 vCPU and cost very different amounts. One word (shared/dedicated) on the pricing page explains why.
Shared: Your vCPUs sit on physical cores that other customers also use. When the host is quiet you get close to a full core. When several machines get busy at once, your process waits. Cheap, and fine for most workloads.
Dedicated: The cores are reserved for your VM only. Same clock speed as a shared core, so it is not faster. It stays the same speed when the host gets busy. Usually costs two to three times more.
Full post with the sysbench test and which workloads need which plan: https://www.doprax.com/tutorial/shared-vcpu-vs-dedicated-vcpu-explained
OVH is increasing prices starting in October. Just like with Hetzner’s earlier price increase, the reason are rising RAM prices. Whether (and to what extent) it will affect your Doprax costs will depend on the instances you're using. We will soon follow up with more information.
Full blog post: https://www.doprax.com/blog/ovhcloud-is-raising-prices-up-to-87-percent
Full blog post: https://www.doprax.com/blog/ovhcloud-is-raising-prices-up-to-87-percent
Two exciting news:
1. You can now resize ProVM if you need more RAM, storage, or CPU. To resize your VM, click on the “Resize” tab, and pick your desired configuration.
2. You can add up to 5 new IPs to your VM! To add a new IP, navigate to the “Network” tab, and click “Add IP”.
1. You can now resize ProVM if you need more RAM, storage, or CPU. To resize your VM, click on the “Resize” tab, and pick your desired configuration.
2. You can add up to 5 new IPs to your VM! To add a new IP, navigate to the “Network” tab, and click “Add IP”.
VPN apps collect data about their users. That’s because in addition to offering a tunnel, they also have an account system. That system belongs to the VPN provider, and it can store all kinds of data about you.
You can remove that risk entirely by self-hosting your VPN.
Read more in our blog: https://www.doprax.com/blog/the-data-your-vpn-app-collects-can-become-a-liability
You can remove that risk entirely by self-hosting your VPN.
Read more in our blog: https://www.doprax.com/blog/the-data-your-vpn-app-collects-can-become-a-liability
We’ve reactivated the “pay by card” payment method!
You can now pay with credit card, Google Pay, or Apple Pay. Check out your dashboard’s “add credit” tab: https://www.doprax.com/v/billing-new/top-up/
You can now pay with credit card, Google Pay, or Apple Pay. Check out your dashboard’s “add credit” tab: https://www.doprax.com/v/billing-new/top-up/
As we mentioned previously, ProVM now supports up to 5 additional IPs per VM, which means separate services, isolated clients, and blue-green deploys all fit on one machine. Fewer VMs to patch, same separation where it counts.
What you can do with additional IPs, full blog post:
https://www.doprax.com/blog/five-ips-one-vm-a-lot-of-room-to-move
What you can do with additional IPs, full blog post:
https://www.doprax.com/blog/five-ips-one-vm-a-lot-of-room-to-move
When the network is unstable, your service or app will sometimes be offline or unavailable.
Five fixes to keep your app online, each with the exact command:
→ Cut external calls, self-host what you can
→ Cache the rest, serve the last good response when upstream dies
→ Short timeouts, not "wait forever"
→ Retry with backoff
→ TCP keepalives + auto-reconnect
~20 min, any Ubuntu VM.
Full tutorial: https://www.doprax.com/tutorial/keep-your-app-reachable-unstable-network
Five fixes to keep your app online, each with the exact command:
→ Cut external calls, self-host what you can
→ Cache the rest, serve the last good response when upstream dies
→ Short timeouts, not "wait forever"
→ Retry with backoff
→ TCP keepalives + auto-reconnect
~20 min, any Ubuntu VM.
Full tutorial: https://www.doprax.com/tutorial/keep-your-app-reachable-unstable-network
Not sure which port to run your service on?
Quick answer, with the reasons:
→ Use 443 (HTTPS). It works from almost every network.
→ Use TCP, not UDP. TCP resends lost data on a bad connection; UDP drops it.
→ Need UDP or an unusual port? Pick a region where all ports are open.
→ Test before you trust it: nc -zv YOUR_SERVER_IP 443
A running service on a blocked port looks exactly like a dead one. Here's how to tell them apart:
https://www.doprax.com/docs/which-ports-protocols-survive-bad-network
Quick answer, with the reasons:
→ Use 443 (HTTPS). It works from almost every network.
→ Use TCP, not UDP. TCP resends lost data on a bad connection; UDP drops it.
→ Need UDP or an unusual port? Pick a region where all ports are open.
→ Test before you trust it: nc -zv YOUR_SERVER_IP 443
A running service on a blocked port looks exactly like a dead one. Here's how to tell them apart:
https://www.doprax.com/docs/which-ports-protocols-survive-bad-network
IPv4 addresses are limited, so providers reuse them. If a past owner sent spam, attacked other servers, or ran an open proxy, the address was added to blocklists. When the IP moves to you, that record often moves with it. This is called a dirty IP.
A dirty IP can:
→ send your email to spam
→ get your API calls and webhooks blocked
→ stop visitors on some corporate networks from reaching your site
The good news: you can check an IP in two minutes, and you can fix most cases. The best time to check is right after you create a server, before you deploy anything.
In the full post: how IPs get dirty, how to check yours, and five steps to take if you have one.
Read it here: https://www.doprax.com/blog/what-is-a-dirty-ip
A dirty IP can:
→ send your email to spam
→ get your API calls and webhooks blocked
→ stop visitors on some corporate networks from reaching your site
The good news: you can check an IP in two minutes, and you can fix most cases. The best time to check is right after you create a server, before you deploy anything.
In the full post: how IPs get dirty, how to check yours, and five steps to take if you have one.
Read it here: https://www.doprax.com/blog/what-is-a-dirty-ip