Network services resource limits
Last updated 2025-01-28
This guide details Fastly resource limits for network services products and features based on your account type. It summarizes the implications of exceeding those limits. Limits for these products and features may be increased or adjusted by other offerings (including packaged offerings) you have purchased.
Cache limits
The cache limits for your account depend on when you became a Fastly customer.
Account created on or after June 17, 2020
If you created your account on or after June 17, 2020, the following cache limits apply.
| Item | Limit | Implications |
|---|---|---|
| Cache file size (with Segmented Caching enabled) | unlimited | None |
| Cache file size (without Segmented Caching enabled) | 20MB | Exceeding this limit when trying to cache a file results in a 503 Response object too large error. |
Account created prior to June 17, 2020
If you created your account prior to June 17, 2020, the following cache limits apply.
| Item | Limit | Implications |
|---|---|---|
| Cache file size (with Segmented Caching enabled) | unlimited | None |
| Cache file size (with streaming miss and without Segmented Caching enabled) | 5GB | Exceeding this limit when trying to cache a file results in a 503 Response object too large error unless Segmented Caching is enabled. |
| Cache file size (without streaming miss and without Segmented Caching enabled) | 2GB | Exceeding this limit when trying to cache a file results in a 503 Response object too large error unless Segmented Caching is enabled. |
Rate and time limits
| Item | Limit | Implications |
|---|---|---|
| API rate limit | 1000 requests/hour | Exceeding this limit results in a Too many requests error. The limit is applied to the authenticated user making the request. See API rate limiting for more info. |
| TLS connections limit | 10 minutes | Exceeding this limit results in a 502 gateway timeout error. |
Request and response limits
| Item | Limit | Implications |
|---|---|---|
| URL size | 8KB | Exceeding the limit results in a 414 URI Too Long error. |
| Cookie size | 32KB | Exceeding the limit results in Fastly stripping the cookie and setting req.http.Fastly-Cookie-Overflow = "1". |
| Maximum response size for tarpitting | 4KB | Exceeding the limit results in no error. Tarpitting isn't applied to responses that exceed the limit. |
| Request header size | 128KB | Depending on the circumstances, exceeding the limit can result in Fastly closing the client connection abruptly, or it can result in the client either receiving a 502 Gateway Error response with an I/O error in the body, or receiving a 503 Service Unavailable response with a Header overflow error in the body. |
| Response header size | 128KB | Exceeding the limit results in a 503 backend read error. Read our developer information about errors for more info. |
| Request header count | 96 | Exceeding the limit results in a Header overflow error. A small portion of this limit is reserved for internal Fastly use, making the practical limit closer to 85. |
| Response header count | 96 | Exceeding the limit results in a Header overflow error. A small portion of this limit is reserved for internal Fastly use, making the practical limit closer to 85. |
req.body size | 8KB | Exceeding the limit results in the req.body variable being blank. Request body payload is available in req.body only for payloads smaller than 8KB. req.postbody is an alias for req.body. |
| Surrogate key size | 1KB | Exceeding the limit results in purging API failures stating "surrogate key too long, must be less than 1024 bytes." Any keys that exceed the limit will be dropped instead of truncated. |
| Surrogate key header size | 16KB | Exceeding the limit results in no error and any keys past the one that exceeds the limit will be dropped. |
| Idle client reuse time | 10 min | Client connections cannot be used after this period of idle and must be reestablished. |
Service, domain, and origin limits
| Item | Limit | Implications |
|---|---|---|
| Services total per account | 10 | Exceeding this limit results in an Exceeding max_total_services error. Contact support to discuss raising this limit. |
| Origins per service | 5 | Exceeding this limit results in an Exceeding max_backends error. Contact support to discuss raising this limit. |
| Domains per service | 20 | Exceeding this limit results in an Exceeding max number of domains error. Contact support to discuss raising this limit. |
| Connections per service | 200 | Exceeding this limit results in an Error 503 backend.max_conn reached error. You can increase this limit as high as 1000 by updating the backend connection setting to limit the connections a single Fastly cache server will make to a specific origin server. |
VCL and configuration limits
| Item | Limit | Implications |
|---|---|---|
| Custom VCL file size | 1MB | Exceeding the limit results in a Content too long error. |
| Maximum VCL file size | 3MB | Exceeding the limit results in a VCL is too long error. |
| Varnish restart limit | 3 restarts | Exceeding the limit results in a Service Unavailable error. This limit exists to prevent infinite loops. |
| ACL container entries count | 1000 | Exceeding the limit results in an Exceeding max ACL entries error. Contact support to discuss raising this limit. |
| Dictionary items count | 1000 | Exceeding the limit results in an Exceeding max dictionary items error. Contact support to discuss raising this limit. |
| Dictionary item key length | 255 characters | Exceeding the limit results in an Item key is too long error. |
| Dictionary item value length | 8000 characters | Exceeding the limit results in an Item value cannot be greater than error. |
| Log line size | 16KB Deliver, 64KB Compute | Exceeding the limit results in truncated logs that could result in delivery errors. |
| Service chains and hops | 20 hops, 6 unique services | Exceeding the limit results in a Service unavailable error with the Loop detected response reason text. Read our developer information about loop detection for more info. |
| Synthetic response characters | No character limit | Synthetic responses have no character limit, but large responses may trigger an error for the custom VCL file size limit. |
| Vary objects count | 50 soft, 70 hard | Exceeding the soft limit results in no error. Newer variants displace the oldest. Active fetches from backends are limited to 70 variants. Exceeding this hard limit results in a Too many variants response. Once fetches complete, objects will be removed until the soft limit is reached. Image Optimizer transformations count towards the variant limit for the original image. |
Varnish workspace limits
A Varnish workspace is an allotted amount of memory that's dedicated to each request process. It's used as temporary storage for objects like headers and local variables, as well as to execute some VCL functions.
Each request process has 256K of workspace per cache node to work with. Subroutines that run on the delivery node (e.g., recv, hash, deliver, log, and error) have 256K of workspace to use for each individual request process, while the subroutines that run on the fetch node (e.g., miss, hit, pass, and fetch) also have 256K of workspace to use.
You can check how much workspace is being used for each request by using the workspace.bytes_total and workspace.bytes_free variables in any subroutine.
If a request process exceeds the workspace limit, Varnish will return a 503 Service Unavailable or 503 Header Overflow error.
Do not use this form to send sensitive information. If you need assistance, contact support. This form is protected by reCAPTCHA and the Google Privacy Policy and Terms of Service apply.
