Top100VPS
Welcome to Top100VPS · Top VPS deals, reviews, and comparisons — 100 hosts rankedAll Posts

4 GB VPS for n8n: RackNerd vs HostNamaste vs CloudCone

Disclosure: VPS100 is reader-supported and editorially independent. Links on this page go to the provider's official site; we are not paid a commission when you click or buy.

For a small self-hosted n8n VPS, RackNerd and HostNamaste both charge $59.99 per year for 4 GB. RackNerd includes 60 GB of SSD storage and states that the plan renews at the same rate. HostNamaste includes 70 GB of NVMe storage and 8 TB of transfer. CloudCone's $92-per-year plan doubles the disk to 120 GB and includes 10 TB of transfer, but costs about $32 more. None of the three plans includes a dedicated CPU guarantee, so the decision comes down to disk headroom, renewal terms, and measured I/O.

Plan and official page

Price

Published resources

Renewal and billing

Main caveat

RackNerd 4 GB special · terms

$59.99/year

4 GB RAM, 3 vCPU, 60 GB SSD, 7 TB transfer, 1 Gbps port

Annual; RackNerd states the special renews at the same rate while retained

Unmanaged; no included backup and no verified dedicated-vCPU guarantee

HostNamaste KVM 4 GB · terms

$59.99/year

4 GB RAM, 3 vCPU, 70 GB NVMe, 8 TB transfer, AMD EPYC 7402P node, Los Angeles

Annual; the promo page lists November 8, 2026, as the end date, but the cart still showed $59.99 on November 15

Unmanaged; 7-day money-back applies to the first yearly VPS order; daily backups are provided as a courtesy

CloudCone SSD VPS 4 · terms

$92/year

4 GB dedicated RAM, 8 vCPU, 120 GB RAID-10 SSD, 10 TB transfer

Annual; renewal fees may change with at least 45 days' notice

No dedicated CPU stated; CloudCone describes VPS and promotional payments as non-refundable

Prices and availability were checked on November 15, 2026.

RackNerd and HostNamaste cost the same. HostNamaste offers 10 GB more NVMe storage and 1 TB more transfer; RackNerd has the clearer renewal commitment and a five-day cancellation policy. CloudCone costs $32.01 more, doubles RackNerd's disk to 120 GB, and includes 10 TB of transfer. Its page lists 8 vCPU but describes only the RAM as dedicated, not the CPU.

HostNamaste's Diwali page lists November 8, 2026, as the offer's end date, yet its cart still showed the 4 GB plan at $59.99 on November 15. Confirm the price at checkout before ordering. RackNerd's promotion banner says "WHILE AVAILABLE," and its special page does not show a stock counter.

n8n VPS requirements: CPU steal and single-core performance

A vCPU count tells you how many threads the provider exposes, not how fast a single thread will run. A 3-vCPU plan can outperform an 8-vCPU plan on a single-threaded workflow when the first machine has a faster core and less steal. n8n runs on Node.js, and JavaScript execution within each Node.js process is single-threaded. Queue mode adds worker processes, but each worker still needs CPU capacity; it does not turn a slow core into a fast one.

Check `%st` in `top`, then run `vmstat 1` and `mpstat -P ALL 1`. `%st` measures the time the hypervisor took away from your VM. Sustained steal that coincides with slow executions is a reason to move. Run Geekbench 6 single-core or YABS after deployment, and test a representative workflow with hundreds of nodes rather than a hello-world webhook.

HostNamaste identifies the node as an AMD EPYC 7402P. RackNerd does not publish a CPU model on its special page, and CloudCone names Intel Xeon processors without identifying the generation. Treat all three as test-after-deploy purchases. If the node you receive is slow or shows high steal, move before you build production dependencies around it.

Disk I/O and database growth on a 4 GB n8n VPS

n8n writes execution data to its database. The official execution-data guide warns that the database can grow until it runs out of storage, and recommends avoiding unnecessary saves and pruning old executions. Set `EXECUTIONS_DATA_SAVE_ON_SUCCESS=none` to stop saving successful runs, keep `EXECUTIONS_DATA_SAVE_ON_ERROR=all`, set `EXECUTIONS_DATA_SAVE_ON_PROGRESS=false`, and set `EXECUTIONS_DATA_SAVE_MANUAL_EXECUTIONS=false`. Successful execution history will disappear, so retain enough error data for debugging.

SQLite is the default database. It stores data in a single file, and n8n's queue-mode guide says queue mode with SQLite is not recommended; that setup expects Redis and PostgreSQL. If you keep SQLite, pruning does not shrink the file automatically. Set `DB_SQLITE_VACUUM_ON_STARTUP=true` or run `VACUUM` manually, as the database documentation explains.

RackNerd's 60 GB disk is the tightest of the three. HostNamaste provides 70 GB of NVMe storage, and CloudCone provides 120 GB of SSD storage. NVMe can reduce latency, but none of the providers guarantees a minimum IOPS level. Test with a 4K random read/write workload using `fio`, then monitor `iostat -x 1` and the `wa` column in `vmstat 1`.

TOS limits and rate limiting for n8n

RackNerd's terms use a fair-share standard. VPS customers must monitor usage and avoid abusive disk I/O or CPU behavior; RackNerd says customers who need dedicated resources should buy a dedicated server. The terms do not specify a numeric CPU percentage or a fixed burst window. RackNerd also states that refunds are not provided by default.

HostNamaste's terms state that VPS usage is limited by the resources allocated to the plan. They do not specify a numeric CPU cap for VPS plans. HostNamaste offers a 7-day money-back guarantee on the first yearly VPS order and provides daily backups as a courtesy, not a guarantee.

CloudCone's AUP is more specific. It prohibits 100% of a vCPU for longer than 2 hours on public cloud instances, disk IOPS that affect other users on the same hypervisor, and 500 Mbps or more for longer than 2 hours. A runaway workflow can hit that limit even when the average load looks low.

Place Nginx `limit_req` and `limit_conn`, or Cloudflare rate limiting, in front of public webhooks. Use fail2ban, and remove webhook routes that do not need to be public. A rate limit will not fix a poorly designed workflow, but it can stop an internet scanner from turning one bad request into a CPU incident.

Which 4 GB VPS to buy for n8n

Choose RackNerd if the same-rate renewal commitment matters most and 60 GB is enough. Watch the disk first: Docker images, execution data, binaries, logs, and backups all share the same 60 GB. RackNerd requires five days' written notice to cancel before renewal.

Choose HostNamaste if you want 70 GB of NVMe storage and 8 TB of transfer at the same $59.99 price, provided that checkout still honors that price. Its 7-day money-back window on the first yearly VPS order reduces trial risk. Confirm the promotion's end date before ordering.

Choose CloudCone if 120 GB of disk, 10 TB of transfer, and a larger published vCPU count are worth about $32 more per year. The extra disk is useful for n8n execution data and Docker layers, but the plan still lacks a dedicated CPU guarantee, and its renewal price can change.

Before moving production workflows, test CPU steal, single-core performance, and 4K disk latency on the actual node. Then cap execution data, enable pruning, and keep backups outside the provider. None of these plans is a guaranteed high-concurrency n8n platform.

For provider-level network checks, see the RackNerd Looking Glass. For a broader renewal-cost comparison, see the VPS True 3-Year Cost Comparison.

n8n VPS FAQ

How much RAM does a self-hosted n8n VPS need?

4 GB is enough for a small n8n deployment running Docker with a modest PostgreSQL or Redis stack. Execution concurrency, binary data, logs, and database growth can change that requirement quickly. Treat 4 GB as a practical starting point, not a capacity guarantee.

What is the best 4 GB VPS for n8n?

There is no single best plan. RackNerd and HostNamaste both cost $59.99 per year. RackNerd provides 60 GB of SSD storage and a same-rate renewal commitment; HostNamaste provides 70 GB of NVMe storage and 8 TB of transfer, but its promo page lists November 8, 2026, as the end date. CloudCone costs $92 per year and includes 120 GB of storage, 10 TB of transfer, and 8 published vCPUs, but it has no dedicated CPU guarantee and less predictable renewal pricing.

How do I check CPU steal?

Run `top` and watch `%st`, then run `vmstat 1` and `mpstat -P ALL 1`. Sustained steal that coincides with slow executions is a warning sign. Run Geekbench 6 single-core or YABS, and test a representative workflow rather than a trivial one.

Does `EXECUTIONS_DATA_SAVE_ON_SUCCESS=none` delete all execution history?

It stops saving successful executions, while errors can still be saved with `EXECUTIONS_DATA_SAVE_ON_ERROR=all`. Pruning controls how long saved data remains. If you use SQLite, pruning does not automatically shrink the database file; set `DB_SQLITE_VACUUM_ON_STARTUP=true` or run `VACUUM`.

Are there fixed CPU time limits?

RackNerd uses a fair-share standard without a published numeric CPU cap. HostNamaste limits VPS usage to the resources allocated to the plan and does not publish a numeric CPU cap. CloudCone's AUP prohibits 100% of a vCPU for longer than 2 hours on public cloud instances. Read the exact terms for the plan you order.

0 comments

No comments yet

Leave a comment