Limits

Limits are the sizes, counts, time limits and quotas that Sevalla enforces on your services. Each table gives the value, what happens when you reach it, and how to raise the limit or work around it, with a link to the page that explains it. A value marked “(default)” can be raised for your company (or, for companies you own, for your account). To raise it, contact support with the limit, the service and the value you need. Every other value is the same for everyone.

Instance sizes

An instance size sets the CPU and memory of each instance of a process. Pick it per process; every instance of the process gets the same size. See Scaling.

Size	Group	vCPU	Memory
H1	Hobby	0.3	0.3 GB
S1	Standard	0.5	1 GB
S2	Standard	1	2 GB
S3	Standard	2	4 GB
S4	Standard	4	8 GB
S5	Standard	8	16 GB
C1	CPU optimized	1	1 GB
C2	CPU optimized	2	2 GB
C3	CPU optimized	4	4 GB
C4	CPU optimized	8	8 GB
M1	Memory optimized	0.5	2 GB
M2	Memory optimized	1	4 GB
M3	Memory optimized	2	8 GB
M4	Memory optimized	4	16 GB
M5	Memory optimized	8	32 GB

Default size. The dashboard picks H1 for a new application’s web process. The API creates it on S1; change it with Update process.
No GPUs. Every size is CPU only. To run a model on Sevalla, run it on CPU, or call a hosted model API from your application.
The largest size is 8 vCPU or 32 GB. For more, run more instances.
Thread counts. Inside an instance, tools such as nproc report the host’s CPUs, not the size’s. Set worker and thread counts (for example WEB_CONCURRENCY or LLAMA_ARG_THREADS) to match the size.

Processes and instances

Limit	Value	At the limit	Raise it or work around it
Web processes	1 per application	`App cannot have multiple web processes`	Run other servers as background workers with a private port or a TCP proxy
Background workers, cron jobs and jobs	No limit per application
Instances per process	0 to 20 in the dashboard (default). The API accepts 0 to 50	The Instance count slider stops at the limit. Above 50, the API returns `400` with `Request doesn't match the schema`	Contact support to raise the dashboard limit, or use horizontal autoscaling
Autoscaling	Minimum 1 or more, maximum up to the instance limit. Targets 1 to 100 %, steps of 1 to 5 instances, intervals of 1 to 60000 s		See Horizontal autoscaling
Payment method	Creating an application and creating or updating a process need a valid payment method on the company	`No valid payment method found. Add one at https://app.sevalla.com/billing/payment-methods to continue.`	Add a payment method in Billing
Memory	The size’s memory, a hard limit	The instance is stopped with exit code `137` (`OOMKilled`) and restarted	A larger size, or a Memory optimized one. See Troubleshoot runtime errors
CPU	The size’s CPU	The process is throttled, not stopped	A larger size, a CPU optimized one, or more instances
Local filesystem	Up to 20 GB per instance, wiped on every restart and deployment	The instance is stopped and restarted	Keep files on a disk or in object storage
Shutdown	`SIGTERM` 5 s after an instance starts stopping; stopped 30 s after it began	In-flight work is cut off	Handle `SIGTERM`, and start long commands with `exec`. See How processes run

HTTP requests

Requests to an application’s domains pass through Cloudflare, then Sevalla’s ingress in the application’s location. See How traffic reaches your application.

Limit	Value	At the limit	Raise it or work around it
Request timeout	Cloudflare waits up to 185 s for the application to answer	Cloudflare returns `524` (`A timeout occurred`)	Answer within the limit: queue long work for a background worker and return at once, or stream the response. See 504 and 524
Response time with hibernation on	The response headers must start within 15 s	Cloudflare returns `504` (`Gateway time-out`)	Send the headers first and stream the body, queue long work for a background worker, or turn hibernation off
Streaming	Responses are not buffered: server-sent events and chunked responses reach the client as they’re written		A stream stays open as long as it keeps sending data. No `X-Accel-Buffering` header is needed
WebSockets	Supported on the web process, through its domains
Session affinity	None: requests spread across all instances of the web process, request by request	Sessions or state held in one instance’s memory or files break with more than 1 instance	Keep sessions and state in a database, such as Redis or Valkey
IP restrictions	No limit on entries		See IP restrictions

Builds and deployments

Limit	Value	At the limit	Raise it or work around it
Whole deployment (clone, build and rollout)	30 minutes (default)	The deployment fails; the previous version keeps running	Keep the build cache on, set a Build path, or deploy a prebuilt Docker image. Contact support to raise it
Rollout of one process	20 minutes	Sevalla rolls that process back and the deployment fails	Make the process start faster, and check its health checks
A job during a deployment	10 minutes by default	`<job name> has timed out after <seconds>s and has been stopped`, and the deployment fails	Set the job’s Max run time, up to 3,600 s. A job still stops at the deployment’s own time limit
Build machine	4 vCPU and 8 GB of memory	A build that needs more memory fails	Build the image elsewhere, for example in CI, and deploy it as a Docker image
Deployments at a time	1 per application. Others wait as Queued		Different applications deploy in parallel
Image platform	`linux/amd64`	Instances of an image built for another platform crash	Build for `linux/amd64`, for example with `docker build --platform linux/amd64`
Build path	Up to 400 characters	`Must be at most 400 characters`

See Time limits and failures and Troubleshoot builds and deployments.

Cron jobs and jobs

Limit	Value	At the limit	Raise it or work around it
Schedule	A 5-field cron expression, so at most one run a minute. UTC unless you set a time zone	`Cron schedule is invalid.`	For work more often than once a minute, use a background worker with its own loop
Overlapping runs	Skip the new run by default	The due run doesn’t start	Pick Replace the running job
Run time of a cron job	No limit by default	With Max run time set, the run stops with `Stopped at its max run time`	See Cron jobs
Retries	A failed cron run is retried once. A failed job isn’t retried
Run history	30 days	Older runs are removed
Instances	1 per run

Disks

Limit	Value	At the limit	Raise it or work around it
Disks	1 per web process or background worker. Cron jobs and jobs can’t have one		Share files between processes through object storage or a database
Sizes	10, 20, 50, 100, 150, 200, 300, 400, 500, 750 and 1000 GB. Above 10 GB needs a payment method
Resizing	Grow only	`A disk cannot be made smaller. Pick a disk resource type of at least <size> GiB.`	Create a new, smaller disk; the old disk’s files are lost
Instances	1 for a process with a disk, and no autoscaling	`A process with a disk can only use manual scaling with at most one instance. Detach the disk to scale it further.`	Scale up with a larger size, or keep files in object storage and scale out
Mount path	Starts with `/`; letters, digits, `_`, `-`, `.` and `/`. Not `/`, `/app`, `/etc`, `/layers`, `/lib`, `/mnt`, `/root`, `/usr/local`, `/var` or `/workspace` themselves	`Invalid path`	Use a folder inside them, such as `/var/lib/data`
Deployments	The process stops before the new instance starts	A short interruption on every deployment

See Persistent storage.

Networking

Limit	Value	At the limit	Raise it or work around it
TCP proxies	3 per application (default)	The dashboard shows `You’ve reached the limit of 3 TCP proxies. Delete one to create another.` The API returns `400` with `Resource limit reached for tcpProxy`	Contact support
Private ports	100 per process (default)	The API returns `400` with `Resource limit reached for privatePort`	Contact support
TCP proxy port	Assigned by Sevalla, from 20000 to 29999 or from 32768 to 65000
Internal connections	Only between services in the same location		Create the services in the same location

See Networking.

Environment variables

Limit	Value	At the limit	Raise it or work around it
Key	1 to 255 characters. Application keys use only letters, digits, `-`, `_`, `.` and `:`	The API returns `400` with `Request doesn't match the schema`
Value	Up to 32,768 characters	The API returns `400` with `Request doesn't match the schema`	Keep larger data in object storage and put its URL in the variable
Number of variables	No limit
References between static site variables	Up to 10 levels

The limits apply to application variables, global variables and static site variables.

Web terminal

Limit	Value	At the limit	Raise it or work around it
Tabs	8 per application	`You can have up to 8 tabs per application.`	Split a tab into panes
Session after the connection drops	Kept 15 minutes	The shell ends	Run long tasks in tmux
Session without input or output	8 hours	The shell ends
Disconnected sessions	20 per user and 10 per process	The oldest one ends
Output kept for a replay	The last 256 KB per session	Older output isn’t replayed
File upload or download	100 MB	`This is over the 100 MB limit.`	Split the file, or fetch it from a URL with `curl` in the terminal

See Web terminal.

Logs and metrics

Limit	Value	At the limit	Raise it or work around it
Time ranges in the dashboard	Custom ranges up to 60 days back, for logs and metrics
Delay	Lines can take up to 2 minutes to appear
Lines in the dashboard view	20,000	Older lines drop off the top	Narrow the time range or the filters, or download the logs
Entries per API request	1 to 5,000, default 1,000	The most recent entries in the range are returned	Ask for shorter time ranges

See Logs.

Databases

Size	CPU	Memory	Storage
DB1	0.25	0.25 GB	1 GB
DB2	0.5	2 GB	5 GB
DB3	1	4 GB	10 GB
DB4	2	8 GB	20 GB
DB5	4	16 GB	40 GB
DB6	8	32 GB	60 GB
DB7	14.5	58 GB	80 GB
DB8	28.5	108.5 GB	90 GB
DB9	58.5	229 GB	100 GB

Limit	Value	At the limit	Raise it or work around it
Connections	Depends on the engine, the size and the location. A DB1 PostgreSQL database in an OCI location allows 100	PostgreSQL refuses new connections with `sorry, too many clients already`; MySQL and MariaDB with `Too many connections`	Use a connection pool in your application, close idle connections, or move to a larger size
Storage	The size’s storage; it can’t be added separately	Writes fail	Move to a larger size before it fills. See Storage
Memory	The size’s memory	The database is stopped and restarted. Redis and Valkey in OCI locations reject writes at 80 % of it	A larger size. MySQL and MariaDB need at least DB2 under production load
Size changes	Up only. DB7 and larger through support		To move down, restore a backup into a new, smaller database
Databases and users per database server	1 database and 1 user		Create another database
Databases per company	No limit
Manual backups	5 per database (Google Cloud), 5 per backup target (OCI)	`Manual backup limit reached, at most 5 manual backups are kept per backup target`	Delete a manual backup first. See Backups
Backup targets (OCI)	5 per database	`Backup target limit reached, a database can have at most 5 backup targets`
Internal connections	Only from services in the same location		Use the external connection

To see a database’s connection limit, connect to it and run the engine’s command:

SHOW max_connections;

SHOW VARIABLES LIKE 'max_connections';

CONFIG GET maxclients

See Databases.

Object storage

Limit	Value	At the limit	Raise it or work around it
Buckets	20 per company (default)	`Object storage limit reached`	Delete a bucket, or contact support
Uploads in the dashboard	300 MB per file, 100 files at a time	`Larger than 300 MB. Upload it with the S3-compatible API.`	Upload over the S3 API
Object size and requests over the S3 API	Cloudflare R2’s limits
Grace period when you roll keys	Up to 168 hours
Location	Fixed when you create the bucket

See Object storage.

Static sites

Limit	Value	At the limit	Raise it or work around it
Static sites	100 per company	`Site creation blocked due to limitations reached for company "<id>"`	Delete sites you don’t use, or contact support
Publish directory	1 GB per deployment	`'<directory>' directory size exceeds 1 GB limit.` and the deployment fails	Keep large media in object storage
Number of files and file size	No limit beyond the 1 GB
Build time	30 minutes per deployment (default)	The deployment fails	Contact support
Build container	4 CPUs and 8 GB of memory
Deployments queued or running	10 per site, production and previews together	`Pending deployment limit reached for static site "<id>"`. A push at the limit starts no deployment	Cancel deployments you don’t need
Free bandwidth	100 GB per billing period, shared by the company’s sites	Beyond it, $0.10 per GB	See Pricing
Free build time	600 minutes per billing period, shared by the company’s sites	Beyond it, $0.05 per minute
Pull request previews	Deleted 30 days after they finish
Edge cache	30 days, or until the next deployment

See Static sites and Build settings.

Custom domains

Limit	Value	At the limit	Raise it or work around it
Applications on H1	No custom domains while the web process uses H1	Adding a domain fails with `Custom domains are not available while the web process uses a hobby instance size. Change the web process to a larger size first.` Moving a web process with a custom domain to H1 fails with `Cannot create hobby process because app has custom domain`	Use S1 or larger for the web process
Custom domains per service	No limit
Domain name	Up to 253 characters, each label up to 63. Not `.ru`, and not starting with `*.`	`Invalid domain name`, `.ru domain names are not supported`	For a wildcard, add the domain and select Wildcard. See Wildcard domains
The same domain twice	A domain can be on one service only, across all of Sevalla	`Domain already exists`	Delete it from the other service first
Sevalla domain name	3 to 63 letters, numbers and hyphens	`Domain name must be at most 63 characters long`

See Custom domains.

Pipelines

Limit	Value	At the limit	Raise it or work around it
Pipelines	50 per company (default)	`Pipeline limit reached`	Contact support
Stages	10 per pipeline, the preview stage included (default)	`Pipeline stage limit reached`	Contact support
Stages per application	1

See Pipelines.

API, webhooks and companies

The REST API counts requests per company, and per company and service for the tiers that act on one service. Every response has x-ratelimit-limit, x-ratelimit-remaining and x-ratelimit-reset headers.

Tier	Limit	What it covers
Service creation	20 per hour	Creating or cloning applications, creating databases, static sites, buckets and load balancers, and changing custom domains
Service mutation	10 per minute per service	Updating, suspending, activating and deleting a service, cache purges, restarts
Deployment	10 per minute per service	Triggering, rolling back and cancelling deployments
General write	30 per minute	Other writes: environment variables, processes, backups, pipelines, webhooks, API keys
Logs	30 per minute	Deployment, runtime and access logs
Read	400 per minute	Every other request

At a limit, the API returns 429 with a retry-after header: wait that many seconds, then retry with backoff. See Rate limits.

Limit	Value	At the limit	Raise it or work around it
Items per page in the API	`limit` 1 to 100, default 25		Page with `offset`. See Pagination
Webhooks	10 per company	`You can only create up to 10 webhooks per company`	One webhook receives every event you pick; filter on the payload
Companies you own	10 (default)	`You have reached the maximum number of companies allowed, if you have a need to create more companies please reach out to our support`	Companies you’re invited to don’t count. Contact support for more

API and CLI

The sizes on this page come from the API. List them to get the current values and the IDs that other operations take.

Task	API	CLI
List instance sizes	List process resource types	`sevalla resources process-types`
List disk sizes	List disk resource types	`sevalla resources disk-types`
List database sizes	List database resource types	`sevalla resources db-types`
List locations	List clusters	`sevalla resources clusters`

In the API, cpu_limit is in thousandths of a vCPU (500 is 0.5 vCPU) and memory_limit and storage_limit are in MB (1000 is 1 GB).

Get started

AI agents

Applications

Pipelines

Databases

Static sites

Object storage

Load balancers

Domains

Git providers

Projects

Developer tools

Company

Billing

Account

Platform

Security and compliance

Support

Instance sizes

Processes and instances

HTTP requests

Builds and deployments

Cron jobs and jobs

Disks

Networking

Environment variables

Web terminal

Logs and metrics

Databases

Object storage

Static sites

Custom domains

Pipelines

API, webhooks and companies

API and CLI

​Instance sizes

​Processes and instances

​HTTP requests

​Builds and deployments

​Cron jobs and jobs

​Disks

​Networking

​Environment variables

​Web terminal

​Logs and metrics

​Databases

​Object storage

​Static sites

​Custom domains

​Pipelines

​API, webhooks and companies

​API and CLI

Instance sizes

Processes and instances

HTTP requests

Builds and deployments

Cron jobs and jobs

Disks

Networking

Environment variables

Web terminal

Logs and metrics

Databases

Object storage

Static sites

Custom domains

Pipelines

API, webhooks and companies

API and CLI