a month ago
Hello Railway Support,
We're preparing a controlled deployment test and need clarification about a resource-limit field returned by Railway's API.
The serviceInstanceLimits response contains:
containers.pidLimit = 1
Could you please clarify:
- What does
pidLimitrepresent, and what are its units? - Does a value of
1correspond to the Linux cgrouppids.maxlimit, or is it a different Railway configuration value? - How can we verify the effective process/thread limit for a Railway container?
- Could this setting prevent an application from launching a supervisor and child processes within the same container?
We're trying to verify compatibility before deployment.
Thank you!
5 Replies
Status changed to Awaiting Railway Response Railway • about 1 month ago
a month ago
containers.pidLimit is the maximum number of processes (PIDs, which on Linux includes threads) a single container can have running at once. It is a plain count, and we apply it directly as the container's cgroup pids.max limit with no multiplier or other conversion.
Your workspace's effective limit is 1000, which is the standard Pro value, and there is no custom override lowering it. A real limit of 1 would allow only the entrypoint process and block any child process, so a supervisor starting workers would fail. At 1000, a supervisor plus child processes in one container is well within the limit. The 1 you saw does not match what is configured for your workspace.
From inside a running container, /sys/fs/cgroup/pids.max shows the limit the container sees, and pids.current in the same directory shows how many are in use.
Status changed to Awaiting User Response chandrika • about 1 month ago
a month ago
Thank you for clarifying.
This is helpful. However, I would like to resolve one discrepancy before proceeding with our deployment test.
You confirmed that our workspace's effective PID limit is 1000, with no custom override.
However, the Railway API returned:
serviceInstanceLimits.containers.pidLimit = 1
for our isolated service stellar-appreciation in the humble-imagination project.
Could you please confirm:
- Why does the API return 1 when the effective configured limit is 1000?
- Is this a reporting/API issue, or could the service actually start with a PID limit of 1?
- Can you confirm that this specific service will receive the 1000 PID limit when deployed?
We are preparing a controlled, voice-disabled test using a supervisor that launches child processes, so we want to avoid a configuration issue.
Thank you for your assistance.
Status changed to Awaiting Railway Response Railway • about 1 month ago
a month ago
The 1000 we gave earlier is your workspace's plan limit, and there is no workspace-level override on it. The value the API returns for a single service instance is read per service, though, and a service-level value takes precedence over the workspace limit, so a 1 on stellar-appreciation is not something we can dismiss as a display issue.
We have not established where that 1 comes from, so we can't confirm yet that this service will start with 1000. If the API is reporting 1 for it, that is the value the container would be given.
To help pin this down, please share the exact GraphQL query you ran and the raw response, including the environment it was run against. Once the service is deployed, /sys/fs/cgroup/pids.max inside the running container shows the limit it actually received.
Status changed to Awaiting User Response Railway • about 1 month ago
a month ago
Hello Railway Support,
Thank you for investigating the unexpected PID limit on our isolated test service.
We've reviewed the exact GraphQL query and our saved diagnostics. Here are the details you requested.
Project: humble-imagination
Project ID: a55affbc-221b-441a-a3d5-47e842ecd62f
Environment: production (within this isolated test project)
Environment ID: 5f4c8719-2564-49f4-826a-edb36b4e0720
Service: stellar-appreciation
Service ID: 5b7e0857-6aec-4283-a3b7-91ea6b7405d4
GraphQL endpoint:
https://backboard.railway.com/graphql/v2
Exact query used, reformatted for readability:
query($serviceId:String!,$environmentId:String!){
serviceInstance(
serviceId:$serviceId,
environmentId:$environmentId
){
serviceName
startCommand
buildCommand
rootDirectory
healthcheckPath
numReplicas
restartPolicyType
}
serviceInstanceLimits(
serviceId:$serviceId,
environmentId:$environmentId
)
environment(id:$environmentId){
id
config(decryptVariables:false)
}
}Query variables:
{
"serviceId": "5b7e0857-6aec-4283-a3b7-91ea6b7405d4",
"environmentId": "5f4c8719-2564-49f4-826a-edb36b4e0720"
}Request timestamp: 2026-10-11T05:34:10.480Z (UTC)
Our saved diagnostic confirms the following resource structure:
- containers.cpu: 2
- containers.memoryBytes: 2000000000
- containers.pidLimit: numeric field
The original raw HTTP/GraphQL response was not preserved. A subsequent operator checkpoint records the observed pidLimit value as 1, but we do not want to misrepresent that checkpoint as a captured raw response.
Before creating another temporary credential to repeat the query, could your team investigate the configuration for this specific service and determine:
- Why the service-specific PID limit appears as 1 despite the workspace limit of 1000?
- Whether a service-level override is responsible and how it can be corrected through a supported mechanism.
- Whether the effective limit can be confirmed through Railway's API before deployment.
- Whether Railway support needs to correct the configuration internally.
The service remains offline. We have not authorized deployment or configuration changes.
If a fresh raw response is essential, please confirm, and we can arrange a separate read-only capture.
Thank you for your assistance.
Status changed to Awaiting Railway Response Railway • about 1 month ago
a month ago
We queried serviceInstanceLimits directly for stellar-appreciation in the production environment you identified, and it currently returns containers.pidLimit: 1000. No PID limit correction is indicated by that result, and you don’t need to create another credential for us to verify the current API value.
Without the original raw response, we can’t determine why the earlier checkpoint recorded 1. Our previous reply was too definitive about the API value guaranteeing the eventual container limit: deployment configuration also factors into that, and /sys/fs/cgroup/pids.max confirms the enforced limit once a container is running.
We haven’t deployed the service or changed its configuration.
Status changed to Awaiting User Response Railway • about 1 month ago
a month ago
This thread has been marked as solved automatically due to a lack of recent activity. Please re-open this thread or create a new one if you require further assistance. Thank you!
Status changed to Solved Railway • about 1 month ago