刚发的自建 VPN 教程，可能有些兄弟还没完全搞懂，这里再补充一种更通用的机房 IP 自建方法，顺便推荐几个好用的 VPS 服务商。 上一条讲的是用 AmneziaVPN 自动部署，这次说的是更自由的手动搭建，协议、客户端、分流规则都能自己选。 其实没那么复杂： 1、先买 VPS：Vultr、DMIT、Hetzner、DigitalOcean、搬瓦工都可以。选 Ubuntu 系统，个人使用 1 核 1G 基本够起步。 2、连接服务器：Windows 打开 PowerShell，输入 ssh root@服务器IP，登录自己的 VPS。 3、部署节点：推荐从 Xray + VLESS Reality 入手，也可以选择 sing-box、Shadowsocks、Trojan 等方案。不会配置就让 AI 根据官方文档生成安装指令、服务端配置和客户端连接信息。 4、连接设备：服务器部署成功、端口放行后，Windows 用 v2rayN，iPhone 用 Shadowrocket 等兼容客户端，导入节点即可。 5、检查 IP：打开 link VPS IP。 这样就有了一条自己管理的机房 IP 节点。 VPS 服务商简单推荐几个： Vultr：机房选择多，后台操作方便。 DMIT：有针对特定跨境线路优化的套餐。 Hetzner：配置性价比高，适合预算有限的人。 DigitalOcean：文档完整，新手容易上手。 搬瓦工：有 CN2 GIA 等线路产品，适合关注中国方向网络质量的人。 OVHcloud：服务器产品丰富，也比较重视 DDoS 防护。 最后说几个重点： ① 机房 IP 不等于住宅 IP，自己搭的节点也不会自动变成住宅出口。 ② WireGuard、OpenVPN 属于 VPN 隧道路线；VLESS、Trojan、Shadowsocks 属于代理路线，工作原理和客户端都不一样。 ③ 不同协议没有绝对的速度排名。服务器线路、延迟、丢包和晚高峰拥塞，往往更影响实际体验。 ④ 买 VPS 别只看价格，还要看公网 IPv4、流量限制、机房位置和实际网络质量。 简单理解：VPS 决定你的服务器和出口，协议决定你怎么连接，客户端决定你怎么使用。 普通机房节点就是这么搭的，不需要额外购买 ISP 或住宅代理。
DigitalOcean on X
The largest group in reception, 28 people, recommends DigitalOcean to others as the cheap, simple host for a small app or a droplet. The keenest complaint is security advisories landing after people had already patched.
First written 27 Oct · what changed · from 149 posts by 123 people
Tracking posts that say “digitalocean”
Track this for your marketPosts per day, last 30 days
as of Nov 3, 11:01 UTCWhat changed
as of 27 Oct · checked once a monthFirst written 27 Oct. When what people say changes, it shows here.
Tone· 79 posts from 70 people· written 27 Oct
How do people feel about DigitalOcean?
Recommendation outweighs complaint: 28 of the section's 70 people point someone else at DigitalOcean as a host that just works, and 16 praise it outright. The disappointed are 7 people, level with the 7 still weighing it up.
The largest group is practical rather than rapturous: they tell other people to use it — for a self-hosted VPN box, for a small app, for a $10/month droplet — and the reasons repeat. Complete docs and easy for beginners, cheap next to managed services, a console one user rates far better than Hetzner's. A founder running Managed Postgres at $15 a month gives "boring reasons": the app already lives there and backups and restores are handled.
A smaller, louder group is enthusiastic: 16 people call it underrated, "the goat rn", or hard to beat on "balance of performance and pricing", and a couple credit support for replying within minutes. 14 people are impressed by the newer agent and AI products — Managed Agents, Agent Droplets, Signals and Simulations, the Inference Router, 90% off cached tokens — with one investor arguing the market sold it as a neocloud in the week it shipped its most complete agent stack.
The negativity is scattered rather than organised: it is too expensive for some, security warnings came a day late, and the VPS offer is no longer "stupid simple".
- Recommends it as a dependable host33
- Loves it17
- Impressed by the new agent and AI products14
- Curious but not yet convinced8
- Disappointed7
See 5 of 79 postsHide the posts
Don't get doublefucked. Get a digitalocean droplet, tell Opus harden this and also install fail2ban. enjoy no surprise $5000 bill.
$DOCN fell 14.8% yesterday to $109. $CRWV fell 15.5%. $NBIS fell 10%. The market sold DigitalOcean as a neocloud. In the same week, it shipped the most complete agentic AI product stack I've seen from a cloud this size. Here's what I think the market is missing. AI agents need far more than GPUs. An agent that fixes a bug or books a meeting needs a sandbox to run in, memory, storage, databases, tools and web access. That is boring cloud plumbing. DigitalOcean already sells it to 680,000 customers. The last four weeks of press releases: > Sep 22: Managed Agents. Runtime, tool access and AI models on one cloud > Oct 1: Agent Droplets. Everything an agent needs for $50/month (solo) or $200/month (team), 16,000+ tools included > Oct 12: Agent Learning and Data Services. Records what an agent does, flags where it fails, tests the fix before customers see it > Oct 13: Bright Data partnership. Agents can search and pull live web data without a separate account Plus a named customer. Amplitude runs its agent product, Wave, on DigitalOcean Managed Agents. The Q2 numbers already point this way: > AI customer revenue run rate: $234M, up 212% > Inference (running AI models, not training them): up ~800% > 85% of AI revenue comes from inference and core cloud, not raw GPU rental > ~70% of AI customers spending $100K+ a year also buy core cloud > Contracted backlog: $894M, 12x a year ago AI customers already bring their CPU, storage and database spend with them. Agents should make that pull stronger. On the neocloud question, the two balance sheets look nothing alike. CoreWeave's bonds yield ~15%. DigitalOcean is profitable (~39% EBITDA margin), is free cash flow positive, and carries net debt of about 0.7x a year's EBITDA. Yesterday it still fell harder than Nebius. However: > None of the agent products has disclosed revenue. Most are in preview, which the company says is not supported for production workloads > Gross margin fell from 61% to 55% in a year, because data centers open before the revenue arrives > $2.76B of data center leases are signed but have not started > The top 25 customers went from 9% to 20% of revenue in a year > At ~7.7x 2027 revenue it is still not very cheap. And yesterday showed the market won't separate it from the group yet For the stock to work from here, I see three paths: > Nearest: Q3 on Nov 4. Revenue at or above $307M and a Q4 guide of ~$327M+ (35% growth). Oppenheimer raised its target to $220 yesterday expecting exactly that > Structural: the agent pull shows up in the numbers. Gross margin stops falling and existing customers spend more each year. That's when the market stops pricing it like a GPU landlord > Moonshot: it becomes the default agent cloud for AI startups. Its ~155 MW of committed capacity could support ~$3.4B of run rate revenue at what its general cloud earns per MW. That's 3x today What I'm watching on Nov 4: > Any usage number for Managed Agents or Agent Droplets > Gross margin at 55% or better > 2027 reaffirmed at 50%+ growth, with a capex and free cash flow plan > Named customers beyond Amplitude I think this will be one of the winners of the Agentic AI thesis. Enterprise customers are going to flock to Digital Ocean to run their agentic workloads/orchestrations. Yesterday, it closed at the 200SMA 👀. Watching for a bounce there.
pls try @digitalocean its underrated
yeah, i hate how digitalocean is trying to be a neocloud, and now their vps offering isn’t stupid simple anymore like it used to be 10 years ago
5 of the 79 posts behind this answer, one per person. The other 74, and every new one as it’s posted, come with tracking. Get early access
Pain points· 41 posts from 33 people· written 27 Oct
What do people complain about with DigitalOcean?
Late security warnings are the largest complaint, from 8 people, ahead of platform reliability at 6. Smaller themes trail well behind: a couple of people each on price, static IPs, blocked IP ranges and AI features that don't behave as expected.
The late-advisory complaints cluster in the last week of October, around the PHP and then Node CVE waves: warnings that arrived hours or a day after people had already started patching, one that never mentioned App Platform, and one poster noting "the patch beat your email". 4 people then describe the cleanup as theirs — upgrading Ubuntu in place, moving PHP to a patched build, moving Node to 22.23.4 or 24.22.1.
Reliability complaints are concrete: a droplet list that times out or redirects to a marketing page, SSH dropping mid-patch, a fresh dry-run droplet whose IP can't be seen, a card step that returned success while failing. 3 people describe lockouts — 2FA broken on the day they urgently needed to patch, a ticket chased after two hours.
Other themes are thin and varied: 3 people find App Platform behind on PHP runtimes; others flag a $25/month static IP per app, an IP that can't be reserved so all DNS has to change, and an inference router whose "budget" setting rarely picks cheap models.
- Security advisories arrive late8
- Platform reliability problems8
- Patching your server is left to you5
- App Platform lags on PHP versions4
- Account lockouts and slow support3
- Performance claims need a grain of salt3
See 5 of 41 postsHide the posts
Bit late don't you think, @digitalocean? The entire world was burning yesterday and scrambling to patch their servers and apps And you send out a warning the day after? What about the NodeJS CVE? Will you send out another email tomorrow? 😹 Intern sleeping?
Oh no, skill issues at DigitalOcean! Guys this is supposed to show my droplets, not time out or send me to a marketing page. Even if I somehow manage to get through to the droplet I’m trying to manage, after a while I’m getting redirected back to this @digitalocean @DOStatus
@digitalocean emailed about a PHP RCE CVE. My WordPress droplet ran @ubuntu 20.04 + OpenLiteSpeed. Plan: upgrade the OS in place, move PHP to the patched build.
As far as I can tell @digitalocean App Platform has still not updated PHP versions.
do I have any friends at @digitalocean who may see this? I'm trying to restore my account but I don't have access. I have files a ticket but would like to move this along quickly
5 of the 41 posts behind this answer, one per person. The other 36, and every new one as it’s posted, come with tracking. Get early access
Alternatives· 40 posts from 37 people· written 27 Oct
What are the alternatives to DigitalOcean?
Hetzner and Vultr are the names that come up most, from 8 and 7 people, usually as the same kind of cheap VPS lined up beside DigitalOcean. Railway is next at 5, and it is the one people are actively leaving for droplets.
Hetzner and Vultr rarely arrive as criticism: they are listed alongside DigitalOcean in the same breath when someone asks where to put a small app or a self-hosted VPN — "Hetzner or Digitalocean". Hetzner carries the cheaper-European reputation, though one poster notes its prices have climbed too, and the one direct judgement on the consoles goes DigitalOcean's way. Vultr is weighed on regions, latency and cost.
Railway is the more pointed comparison: 5 people, several describing a move off it — an urgent alert that prompted a migration, deploys going "from 6 minutes to 27 seconds", fixes that couldn't be shipped. One person went the other way, moved to Railway for the UX and now regrets it. AWS appears as the complicated, effortful option people are glad to have left; Vercel and Railway get the same overpriced label from one switcher.
Cloudflare shows up as where workloads went next; one consolidation post naming it travelled far but is one person's account.
- Hetzner8
- Vultr8
- Railway7
- AWS4
- Cloudflare3
- Akamai3
See 5 of 40 postsHide the posts
Holy shit the console of @Hetzner_Online is so shit in comparison to @digitalocean
Vultr or DigitalOcean? Start with your route and requirements. 🐦 Compare current regions, latency, transfer allowances and total cost. Identical software can perform differently across networks. What matters most in your shortlist? link
Deploys went from 6 minutes to 27 seconds on DigitalOcean compared to Railway. Easier to iterate, fantastic.
i moved away from digitalocean to railway thinking its better and from a UX, and yes it is but now I immediately regret my decision looking for alternatives as well
Due to recent nodeJS CVE, I have just moved my very old iOS Dev blog link away from DigitalOcean VPS Ghost image, to just static HTML files on Cloudflare 🤞
5 of the 40 posts behind this answer, one per person. The other 35, and every new one as it’s posted, come with tracking. Get early access
Questions· 6 posts· written 27 Oct
What do people ask about DigitalOcean?
The questions are few and practical, from 6 people, split level between what hosting actually saves, whether a droplet is the right thing to pick, and what AI usage really costs.
Nobody asks about features in the abstract. One person wants a number: what percentage was saved by running frontend, backend and Postgres on DigitalOcean rather than splitting them across services, and another asks how to halve a bill using an agent on DigitalOcean.
The droplet question comes from people who don't yet know the product — one asking whether it is the pick, another asking whether droplets are like AWS Lightsail and saying they'll go and research what they are.
The pricing questions are about the newer AI products: whether benchmark numbers were run on the $0.15 input tier or a dedicated deployment, and what resume latency applies before an idle agent is cheaper to keep warm than to suspend.
- How much can I save on hosting?2
- Are droplets the right fit?2
- How is AI usage priced?2
See 4 of 6 postsHide the posts
did you track roughly how much you saved hosting the frontend, backend and Postgres all on DigitalOcean instead of splitting them across separate services? curious about the %
@grok How could we cut this bill in half with Replit, Claude, and Hermes installed on DigitalOcean? Use all known Github repos, agent skills, and open source alternatives.
ah ok, gracias por la respuesta estas estarian en droplets de DigitalOcean ? , no conozco mucho DigitalOcean, voy a investigar que serian. Seria algo parecido a los LightSail de AWS?
Were those benchmark runs served on the $0.15 input tier, or a dedicated deployment?
4 of the 6 posts behind this answer, one per person. The other 2, and every new one as it’s posted, come with tracking. Get early access
Wishes· 3 posts· written 27 Oct
What do people want from DigitalOcean?
Three separate asks, one person each: say "ship this to DO" and get a consistent result, be able to set your own project icons, and see pause-resume latency published with proof that a paused agent lost nothing.
The wishes are few and unrelated, so they read as individual requests rather than a shared demand. One person wants reliability, ergonomics and cost management to line up so that "let's ship this to DO" produces consistent quality — the ask is about predictability as much as the platform.
Another is blunt about a console gap: custom project icons, "it's just silly at this point".
The third is the most technical, and aimed at the agent products: publishing resume latency is good, but what matters alongside it is state integrity — after a pause, being able to show nothing was lost and no action was duplicated, with idempotency as the real test.
- Ship to DigitalOcean and get a consistent result1
- Custom project icons1
- Prove state survives a paused agent1
See the 3 postsHide the posts
Reliability, ergonomics for ai, cost management. I should be able to say "let's ship this to DO" and get consistent quality result.
@digitalocean it's 2026, when are we going to be able to customize our project icons. It's just silly at this point.
Resume latency is a good metric to publish. For agents, what matters alongside it is state integrity: after a pause, can the agent prove nothing was lost and no action was duplicated? Idempotency is the real test.
In their words
How do people talk about DigitalOcean?
The phrases people use most, counted once per person.
- 1“digitalocean droplet”7
- 2“app platform”4
- 3“per token”3
- 4“moving to digitalocean”2
- 5“away from digitalocean”2
- 6“resume latency”2
- 7“nodejs cve”2
- 8“ip address”2
- 9“bare metal”2
Posts are shown the way X shows them. One of them yours? Here’s why it’s shown, and how to take it off.
Also in App hosting and deploy
See all 6 in App hosting and deploy →- Cloudflare+92%The reception is positive and specific: the largest group of 244 people praise tools that just work, with the free tier and protection next.555 posts / 7d
- Railway+69%The conversation is about one long outage and an exposure of encrypted env vars: anger at how it was handled, cancellations, migration to AWS and demands for refunds, alongside users who still call Railway the easiest place to ship and say they are staying.529 posts / 7d
- Vercel—Vercel is a default host people recommend: the largest group praises how fast a project goes live from a repo, plus its free tier.429 posts / 7d
- Netlify−50%Satisfied users recommending Netlify for quick, easy deploys are the largest group in the conversation, with its AI and agent features a smaller, enthusiastic strand.379 posts / 7d
Get every post about DigitalOcean, filtered and labelled
This page is the summary. Tracking gives you the stream behind it for any brand, keyword or @mention: each new pain point, request and question as it’s posted, with the spam and promotions already taken out.