Railway for n8n

Railway sleeps a service only when you turn on Serverless. Railway's guide says: Serverless is a per-service toggle. When enabled, Railway watches the service for activity and stops it after 10 minutes of inactivity. A slept service accrues no compute charges.[S10] It also says: A service that runs 24/7 but only handles traffic a few hours a day still pays for every idle minute. Serverless mode fixes this by putting inactive services to sleep and waking them on the next request.[S9]

The setting lives in the service, not the workspace: Navigate to your service's settings > Deploy > Serverless[S13], then toggle Enable Serverless. A change only applies after a redeploy: Serverless is applied to a container when that container is created. Enabling it on a running service does not affect the container already running; the service must be deployed before it takes effect.[S8]

No Railway page we read says Serverless is off by default in those words. Serverless used to have another name: Note: This feature is formerly called: App-Sleeping[S1]

• It watches outbound traffic: When Serverless is enabled for a service, Railway automatically detects inactivity based on outbound traffic.[S2]
• It sleeps the service after about 5–10 minutes with no outbound traffic (Railway's pages say 5 and 10). The reference page says: Once a service stops sending packets it is considered inactive after 5 minutes. Inactivity is sampled on an interval rather than measured continuously, so in practice a service sleeps somewhere between 5 and 10 minutes after its last outbound traffic.[S3] The guide says 10 minutes (Serverless is a per-service toggle. When enabled, Railway watches the service for activity and stops it after 10 minutes of inactivity. A slept service accrues no compute charges.[S10]).
• It wakes on incoming traffic: A service is woken when it receives traffic from the internet or from another service in the same project through the private network[S5]
• The first call can fail: The first request sent to a slept service may return a 502 Bad Gateway response.[S6]
• A slept service can be de-prioritized: Slept services still consume a slot on Railway's infrastructure, enabling Serverless de-prioritizes your workload and in remote cases, may require a rebuild to re-live the service.[S7]
• An open database connection can keep the service awake. Railway lists, among things that can prevent sleep: Keeping active database connections open, such as a database connection pooler.[S4]

Webhooks: a webhook call is incoming traffic, so it wakes a slept service (A service is woken when it receives traffic from the internet or from another service in the same project through the private network[S5]). The first call may get a 502 (The first request sent to a slept service may return a 502 Bad Gateway response.[S6]). Railway lists Webhook receivers that fire a few times a day[S11a] as a good fit for Serverless.

Schedules: Railway's docs list incoming traffic as the wake-up trigger and mention no timer wake-up (A service is woken when it receives traffic from the internet or from another service in the same project through the private network[S5]). We have not tested n8n's Schedule Trigger on a slept service. Our advice: keep Serverless off if your n8n runs schedules.

Postgres: Railway's n8n template sets up n8n with Postgres. Railway lists Workers that hold database connections or poll queues (they never sleep anyway)[S11b] and Databases[S11c] as bad fits, and an open database connection can prevent sleep (Keeping active database connections open, such as a database connection pooler.[S4]). So with Postgres, Serverless may never put n8n to sleep.

n8n also has a queue mode for scaling. Railway's bad-fit list covers workers that poll queues, and says they never sleep anyway.

n8n's Schedule Trigger docs start with: If a workflow uses the Schedule node as a trigger, make sure that you save and publish the workflow.[T1]

The node's Settings tab has options controlling what happens when an execution misses its scheduled time, for example because the instance was down[T2a]. The If Execution Is Missed setting has three choices. The default, Don't Run Missed Executions, is described as: Discard missed executions and resume on schedule.[T2b] The others are Run the Most Recent Missed Execution and Run the Most Recent Missed Execution Per Rule. A grace period sets how late a run may start before it counts as missed. Set to 0, it uses the instance setting N8N_SCHEDULER_MISFIRE_GRACE, 60 seconds by default.

Which n8n versions this covers: These options are available from n8n 2.36, on Schedule Trigger nodes added from that version on. They take effect only when the instance runs the durable scheduler, which is on by default from n8n 2.45.0. If you set N8N_SCHEDULER_ENABLED=false, Schedule Trigger nodes use the in-memory scheduler, which never runs missed executions.[T2c]

n8n's durable scheduler page describes both schedulers. In-memory: When an instance stops, its in-memory timers go with it. n8n skips any run whose time passed during the downtime rather than catching it up.[T3a] Durable: A run whose time passed while the instance was down still fires late when the instance comes back, as long as it's within its grace period; beyond that, the trigger's misfire policy decides what happens to it.[T3b]

Our reading, not tested: a slept Railway service is stopped (Serverless is a per-service toggle. When enabled, Railway watches the service for activity and stops it after 10 minutes of inactivity. A slept service accrues no compute charges.[S10]) and wakes only on incoming traffic (A service is woken when it receives traffic from the internet or from another service in the same project through the private network[S5]), so a schedule cannot wake it. n8n's docs describe runs missed while the instance was down, not Railway's sleep. If n8n treats a slept period the same way, the If Execution Is Missed setting would decide what happens to those runs. We have not tested this on Railway. To see how many runs a schedule makes, read what n8n costs by executions.

Webhooks: n8n's Webhook docs say the production URL works only while the workflow is published. Their table row for the Production URL reads Publish the workflow[W1a] and Until workflow is unpublished[W1b]. Railway's n8n guide adds: n8n needs to know its public URL to generate correct webhook endpoints.[N4] Set-up steps are in n8n setup problems, fixed at the source. On a slept service, the incoming call wakes it (A service is woken when it receives traffic from the internet or from another service in the same project through the private network[S5]) and the first request may get a 502 (The first request sent to a slept service may return a 502 Bad Gateway response.[S6]). We have not checked whether a given webhook sender retries after a 502.

• A compute hard limit you set: Hitting the Compute hard limit takes your workloads offline but leaves the agent available.[L1] The minimum amount you can specify as the hard limit is $10.[L2]
• Billing problems. Railway's FAQ lists these reasons under "My services were stopped": Usage limits reached … Trial credits exhausted … Failed payment … Unpaid invoice[L4].
• Cancelling: When you cancel your subscription, Railway will stop all deployments in your workspace to prevent further charges.[L5]
• The trial: After 30 days passes or $5 is spent, the free trial reverts to the Free plan, which provides $1 of free credit per month.[L7] On an unverified account: services on the Limited Trial have restricted outbound network access and only a limited set of ports are available.[L6] See free n8n hosting for other free options.

Hobby is $5/mo and includes $5 of usage (the price dataset, checked Oct 19, 2026). Railway's docs say: The Hobby plan includes $5 of resource usage per month.[P2] An always-on n8n pays for every minute it runs. You pay for the RAM and CPU your running service uses, by the minute, even when idle.

If usage goes over the included amount, you pay the usage total: If your resource usage is $7, your total bill for the cycle will be $7 ($5 subscription fee + $2 of usage)[P3]. If it stays under, you still pay the subscription: Even if you do not use the $5 in usage (CPU, Memory, egress), you always pay the $5 subscription fee.[P4]

We cannot give a dollar figure for n8n, because n8n's RAM and CPU use on Railway is not verified. Find your own RAM figure in How much RAM n8n needs. Fixed-price plans are in the cheapest way to host n8n and n8n hosting under $5/mo.

n8n's data lives on the volume: n8n stores workflow definitions, credentials, and execution data in /home/node/.n8n by default. Without a volume, this data is lost on every deployment.[N2] Back it up the way n8n backups explains.

Railway's own advice for a new service: Allow it to run for one week[F1], then Check your Estimated Usage in the Usage Section of your Workspace settings[F2].