Http error 555

Does the rise of serverless mean we need a new HTTP status code?

The team at Oracle think so. They’ve submitted a draft specification to the HTTP Working Group, defining a new HTTP status code (initially suggesting 555) to be used for server-side errors caused by user-supplied resources.

[Note: I’m going to use 555 to refer to the new proposed code everywhere here, but this is not standardized, even if it is standardized in future it will probably use a different code, and you 100% should not start building anything that uses this code for real anywhere. Nobody needs another 418 I’m A Teapot battle.]

Anyway, let’s talk about what this means, and whether it’s a good idea.

Status codes: a refresher

First let’s recap the background. Status codes are 3 digit codes, included in every response from an HTTP server, which summarize the result of a request.

There’s common examples you’ll have heard of like 404 (the resource you requested could not be found) or 200 (your request was successful, and this response represents the result). There’s then a long list of less common examples, like 302 (the resource you requested is a temporarily stored elsewhere, you should go there instead), 410 (the resource you requested was here, but now it’s gone, and there’s no new address available), or 100 (yes, please continue sending the rest of the request you’re already sending).

And many more. They’re categorized into a few classes:

1XX: Information

These are provisional responses, which typically don’t fit into the simple HTTP request/response flow, and describe unusual behaviors like interim responses or switching the connection to a different protocol.

2XX: Success

The request you asked for was successful in some way. Perhaps you’re getting the resource you asked for (200), the server has accepted and started asynchronously processing your operation (202), or your request was successful but the server doesn’t have any data about it for you (204).

3XX: Redirection

Your request is valid, but the response you requested requires you to take some action elsewhere. Perhaps the resource you want is currently only available elsewhere (301/302/307/308), or your request indicated that your cache is up to date, and already has the data you need (304).

4XX: Client error

You, the client sending the request, have done something wrong. Maybe the server can’t understand you at all (400), you’re not authenticated for the thing you’re asking for (401), there was a conflict between your request and the current state of the server (409), or you’ve made too many requests recently and the server is rate limiting you (429).

5XX: Server error

Your request seems valid, but the server receiving your request can’t deal with it for some reason. The entire service might be unavailable (503), an upstream server the server wants to pass your request too might have failed (502), or server might have broken in some way it can’t explain (500).

These classes are well defined and widely understood nowadays, and very unlikely to change in future. It is however possible and likely that new individual codes within each class will be needed, and there’s some details of how that should work in RFC 7231.

These were designed to be extensible from early on, and any client that receives an unrecognized YXX status is required to treat it a Y00. For example, clients that receive 555 responses and don’t understand them are required to treat them as 500 responses. Whether they do in practice of course is a separate question…

Errors as a service

Back to the proposed status code 555. Why do the Oracle team want it?

Oracle are building a service called Oracle Rest Data Services, a developer platform designed to let you quickly create REST APIs for your Oracle databases. You define the endpoints and the SQL, and they generate the API (I’ll carefully avoid discussing whether this is a good idea).

They’re in good company here — the developer market for cloud-hosted software platforms has exploded in the last couple of years, with a massive range of serverless providers and related tools appearing everywhere, from AWS Lambda to Firebase Cloud Functions to Cloudflare Workers to OpenFaaS.

In each case, developer platforms like these hide all server concerns and mechanics from you, and provide you with a clear interface and set of expectations to against which to write code. You provide the core business logic for your API, and they do all the server management & heavy lifting.

Sometimes though, this can go wrong. Your code can fail completely: not just fail to run an operation and return an explicit error status, but entirely fail to fulfill the core expectations required by the platform. Perhaps your SQL statement for Oracle RDS is fundamentally broken, or your code crashes before registering the handler that Lambda expects, or calls the callback with gibberish, or your worker totally runs out of memory.

In these cases the platform needs to tell the client sending the request that something has gone wrong. That is definitely some kind of 5xx error. It’s not the client’s fault, and something has gone wrong on the server end. But which 5xx error?

Here’s the list of standardized codes we have to pick from:

• 500 Internal Server Error
• 501 Not Implemented
• 502 Bad Gateway
• 503 Service Unavailable
• 504 Gateway Timeout
• 505 HTTP Version Not Supported
• 506 Variant Also Negotiates
• 507 Insufficient Storage (WebDAV)
• 508 Loop Detected (WebDAV)
• 510 Not Extended
• 511 Network Authentication Required

500, the maximally unspecific option, is the only real candidate.

Unfortunately, the platform can also fail in unexpected ways, and the most appropriate status code for those is also 500.

What Oracle are arguing is that these two cases (the platform failing and the platform customer’s logic failing) are generic & widely relevant cases, and that they are semantically different from one another in an important way. They want to differentiate these cases by status code. It’s a good idea to standardize a status code to do so if the specific case is often going to affect clients’ interpretation of the response, and it’s a widely relevant distinction, so there are many other services who hit equivalent cases and would use the status codes to differentiate them.

The spec itself and the email submitting it have more detail on their reasoning here, and how they propose this works.

So, the big question: do we really need a new status code for this?

Is this widely relevant?

I think there’s a strong argument that these are types of error that are relevant to a huge & growing set of services and clients.

There are many PaaS providers now where this could happen, and they’re increasingly popular. As of 2018, Lambda was running trillions of executions every month. DataDog did an analysis of their customer’s infrastructure in Feb 2020, and half of their AWS users are using Lambda, an adoption rate that increases to 80+% in large and enterprise environments. At the end of 2019, more than 2 million mobile apps were communicating with the Firebase platform every month. Cloudflare launched Workers in 2018, and according to their S-1 filing by mid-2019, more than 20% of all new Cloudflare customers were using it.

These specific platforms won’t last forever, but the running-your-code-within-a-managed-platform model seems likely to be a long-term fixture.

There’s a lot of these platforms around, a lot of services running on them, and a lot of HTTP requests communicating with those services. All of these platforms fail sometimes. Errors from the platforms themselves are a real & widespread issue.

While it’s tough to get hard numbers, it’s also easy to be confident that the code hosted by these platforms often fails too. It’s pretty clear that the both the hosted code errors and platform errors are real cases that are widely relevant to a lot of modern HTTP traffic.

Is this semantically important to differentiate?

This is less clear. Even if these errors do happen widely, do we really need a separate HTTP status code for hosted logic errors and platform errors?

There’s definitely an argument that it’s a pure implementation detail, and the client doesn’t care. The server hit sent an error and something broke. HTTP status codes shouldn’t depend on the infrastructure choices used by the service, they should just tell the client details about their request.

At the same time, there are other existing 5xx codes that explicitly tell us about implementation details of the failing service, when it’s widely useful to do so. 502 and 504 both declare that the service is internally dependent on another server, and the second server has failed (502) or timed out (504), but the server you’re talking to is functioning correctly. Meanwhile 506 tells us that the internal configuration of content negotiation in the server is broken, placing the blame on that specific part of the server’s implementation.

The gateway errors are a pretty similar case to these platform errors, but directing blame at the «which server» level, rather than the «which organization» or «which level of the stack» level that we’re considering here. It’s common that requests go through a CDN or reverse proxy of some sort, and when that fails it’s often useful to know whether it’s the gateway server that has failed, or the actual service behind it, so we have error codes to distinguish those cases. This would be similar.

In practice though, would this really be useful?

The AWS Lambda outage thread above has a nice quote:

Same here too! Getting «Service error.» when I make requests to my functions..Not good aws! I spent a good amount of time thinking it was my mistake since I was working on some of my functions

This is the situation we’re trying to disambiguate. Is the platform failing somehow, or is the hosted code broken?

This is clearly a meaningful distinction for the developers of the service (i.e. the customers of the platform), like the commenter above. When their service starts serving up errors, their understanding of the response and their next steps are completely different depending on which part is failing. Clearer status codes mean fewer sad emojis.

It’s also an extremely important distinction for the platform provider (i.e. AWS/Oracle/Cloudflare/Google/etc). They’d like to be able to monitor the health of their platform. To do so, they’re very interested in failures caused by the platform, but largely uninterested in failures caused by the hosted code within. It’s easier to set up monitoring tools and automation to report on status codes than it is to parse detailed error information from the response itself. It’s also valuable to them because it clarifies the situation to their customers (as in the quote above), and so avoids unnecessary support requests.

Oracle dig into this in their submission:

When such a resource raises an error the only appropriate HTTP status code to use is 500 Internal Server Error. This causes confusion as it looks like there is something wrong with ORDS, where in fact there is only something wrong with the user supplied SQL. Despite explanatory text to clarify this, operators and users very often miss this distinction, and file support issues against ORDS. Further, automated tools that monitor the access log only see the 500 status code, and thus cannot differentiate between ‘real’ 500 errors in ORDS itself that need remediation versus 500 errors in the user supplied REST APIs that probably do not need any remediation.

Still, the developers of a service & the platform hosting the service are not the main clients of a server.

I do think differentiating these two cases is also arguably useful as a client of an API though, uninvolved in the implementation behind it.

This is a debatable point. It is really only relevant to API clients, as a technical audience, rather than browser visitors, but API clients are still important consumers of HTTP responses. For those clients, a platform failing entirely is a meaningfully different case from the service they want to talk to failing. It affects who they should contact to report the issue, how they should categorize it in their own error logs, which status pages to monitor to know when the issue is resolved, and what kind of expectations they can have for the resolution of the issue.

As with gateway errors: when multiple parties are involved in a failing response, it’s useful for HTTP clients to be able to tell who’s fault it is from the outside.

Is this the right solution to the problem?

Ok, let’s take as a given that this is a widespread case that it’s often important to distinguish. Is the 555 status code described the right way to do that?

One alternative would be to distinguish these cases in an HTTP header or response body of a 500 response. That’s not quite as easy to integrate into much automation though, and less visible for something that is (Oracle would argue) an important distinction in the failure you’re receiving. As a platform, if you want your customers to more clearly understand where errors come from, you want it to be obvious.

Unfortunately, there’s one big reason the 555 status code as proposed isn’t viable anyway: for most platforms, it doesn’t make 500 errors any less ambiguous.

The issue is that for many of these platforms it’s possible for hosted code to explicitly return a 500. This is a problem. If 555 is defined to mean «the hosted code crashed», that means that 500 now means either «the hosted code explicitly returned a 500 error» or «the platform crashed». That makes it useless. Users can’t spot platform issues by looking for 500 errors, and similarly platforms can’t monitor their own status by doing so, which means the differentiation is pointless. This is bad.

It’s fixable though. Instead, we can just flip the proposal on its head, and reserve 555 for platform errors, rather than errors in the hosted logic. I.e. if the platform fails in any unknown way, it should return a 555. Platforms just need to watch their monitoring for 555 errors, and developers & API clients can know that 555 errors are always caused by the service’s platform, not the service itself, so everything except 555 is semantically related to the service.

I suspect in Oracle’s case they missed this simply because it’s not relevant to their platform; their hosted code doesn’t appear to be able to directly set the status, just the data, so it can never explicitly return a 500. It’s definitely relevant for other platforms though, from Lambda to Firebase, so without this the spec is probably unusable.

Do we really need a new status code?

Even if we flip this proposal to define a 555 «Unknown Platform Error», given all the above: do we really need this?

It’s hard to definitively answer. I do think there are legitimate arguments for and against, and I don’t think it’s 100% clear cut either way.

The real test is whether the rest of the ecosystem displays any interest. If this is a status code that only Oracle care about, then it really doesn’t need formal standardization. On the other hand, if AWS or other platforms or API clients do start displaying interest, then maybe it’s honestly a widespread and semantically meaningful class of errors. You can debate the theory all you like, but HTTP, like most standards, is intended to be defined by what’s important for real use cases in the wild, not just what one company wants to implement today.

We’ll have to wait and see.

In the meantime, if you want to keep an eye on this and other HTTP developments, subscribe to the IETF HTTP Working Group mailing list for more thrilling specs and debate, or just subscribe to the HTTP Toolkit blog, and I’ll write up the interesting parts.

Originally posted on the HTTP Toolkit blog

Информационные 100 Continue «Продолжить». Этот промежуточный ответ указывает, что запрос успешно
принят и клиент может продолжать присылать запросы либо проигнорировать
этот ответ, если запрос был завершён. Только HTTP/1.1 101 Switching Protocol «Переключение протокола». Этот код присылается в ответ на запрос
клиента, содержащий заголовок Upgrade:, и указывает, что
сервер переключился на протокол, который был указан в заголовке. Эта
возможность позволяет перейти на несовместимую версию протокола и обычно
не используется. Только HTTP/1.1 102 Processing «В обработке». Этот код указывает, что сервер получил запрос и
обрабатывает его, но обработка ещё не завершена. Только HTTP/1.1 103 Early Hints «Ранние подсказки». В ответе сообщаются ресурсы, которые могут быть
загружены заранее, пока сервер будет подготавливать основной ответ.
RFC 8297 (Experimental). Только HTTP/1.1 Успешные 200

OK

«Успешно». Запрос успешно обработан. Что значит «успешно», зависит от
метода HTTP, который был запрошен:

• GET: «ПОЛУЧИТЬ». Запрошенный ресурс был найден и передан в теле
  ответа.
• HEAD: «ЗАГОЛОВОК». Заголовки переданы в ответе.
• POST: «ПОСЫЛКА». Ресурс, описывающий результат действия сервера на
  запрос, передан в теле ответа.
• TRACE: «ОТСЛЕЖИВАТЬ». Тело ответа содержит тело запроса полученного
  сервером.

HTTP/0.9 и выше 201 Created «Создано». Запрос успешно выполнен и в результате был создан ресурс.
Этот код обычно присылается в ответ на запрос PUT «ПОМЕСТИТЬ». HTTP/0.9 и выше 202 Accepted «Принято». Запрос принят, но ещё не обработан. Не поддерживаемо, т.е.,
нет способа с помощью HTTP отправить асинхронный ответ позже, который
будет показывать итог обработки запроса. Это предназначено для случаев,
когда запрос обрабатывается другим процессом или сервером, либо для
пакетной обработки. HTTP/0.9 и выше 203 Non-Authoritative Information «Информация не авторитетна». Этот код ответа означает, что информация,
которая возвращена, была предоставлена не от исходного сервера, а из
какого-нибудь другого источника. Во всех остальных ситуациях более
предпочтителен код ответа 200 OK. HTTP/0.9 и 1.1 204 No Content «Нет содержимого». Нет содержимого для ответа на запрос, но заголовки
ответа, которые могут быть полезны, присылаются. Клиент может
использовать их для обновления кешированных заголовков полученных ранее
для этого ресурса. HTTP/0.9 и выше 205 Reset Content «Сбросить содержимое». Этот код присылается, когда запрос обработан,
чтобы сообщить клиенту, что необходимо сбросить отображение документа,
который прислал этот запрос. Только HTTP/1.1 206 Partial Content «Частичное содержимое». Этот код ответа используется, когда клиент
присылает заголовок диапазона, чтобы выполнить загрузку отдельно, в
несколько потоков. Только HTTP/1.1 Сообщения о перенаправлениях 300 Multiple Choice

«Множественный выбор». Этот код ответа присылается, когда запрос имеет
более чем один из возможных ответов. И User-agent или пользователь
должен выбрать один из ответов. Не существует стандартизированного
способа выбора одного из полученных ответов.

HTTP/1.0 и выше 301 Moved Permanently

«Перемещён на постоянной основе». Этот код ответа значит, что URI
запрашиваемого ресурса был изменён. Возможно, новый URI будет
предоставлен в ответе.

HTTP/0.9 и выше 302 Found

«Найдено». Этот код ответа значит, что запрошенный ресурс
временно изменён. Новые изменения в URI могут быть доступны в
будущем. Таким образом, этот URI, должен быть использован клиентом в
будущих запросах.

HTTP/0.9 и выше 303 See Other «Просмотр других ресурсов». Этот код ответа присылается, чтобы
направлять клиента для получения запрашиваемого ресурса в другой URI с
запросом GET. HTTP/0.9 и 1.1 304 Not Modified «Не модифицировано». Используется для кеширования. Это код ответа
значит, что запрошенный ресурс не был изменён. Таким образом, клиент
может продолжать использовать кешированную версию ответа. HTTP/0.9 и выше 305 Use Proxy «Использовать прокси». Это означает, что запрошенный ресурс должен быть
доступен через прокси. Этот код ответа в основном не поддерживается из
соображений безопасности. Только HTTP/1.1 306 Switch Proxy Больше не использовать. Изначально подразумевалось, что » последующие
запросы должны использовать указанный прокси.» Только HTTP/1.1 307 Temporary Redirect «Временное перенаправление». Сервер отправил этот ответ, чтобы клиент
получил запрошенный ресурс на другой URL-адрес с тем же методом, который
использовал предыдущий запрос. Данный код имеет ту же семантику, что код
ответа 302 Found, за исключением того, что агент
пользователя не должен изменять используемый метод HTTP: если в первом
запросе использовался POST, то во втором запросе также
должен использоваться POST. Только HTTP/1.1 308 Permanent Redirect

«Перенаправление на постоянной основе». Это означает, что ресурс
теперь постоянно находится в другом URI, указанном в заголовке
Location: HTTP Response. Данный код ответа имеет ту же
семантику, что и код ответа 301 Moved Permanently, за
исключением того, что агент пользователя не должен изменять
используемый метод HTTP: если POST использовался в первом
запросе, POST должен использоваться и во втором запросе.

Примечание: Это экспериментальный код ответа,
Спецификация которого в настоящее время находится в черновом виде.

draft-reschke-http-status-308 Клиентские 400 Bad Request «Плохой запрос». Этот ответ означает, что сервер не понимает запрос
из-за неверного синтаксиса. HTTP/0.9 и выше 401 Unauthorized «Неавторизованно». Для получения запрашиваемого ответа нужна
аутентификация. Статус похож на статус 403, но,в этом случае,
аутентификация возможна. HTTP/0.9 и выше 402 Payment Required «Необходима оплата». Этот код ответа зарезервирован для будущего
использования. Первоначальная цель для создания этого кода была в
использовании его для цифровых платёжных систем(на данный момент не
используется). HTTP/0.9 и 1.1 403 Forbidden «Запрещено». У клиента нет прав доступа к содержимому, поэтому сервер
отказывается дать надлежащий ответ. HTTP/0.9 и выше 404 Not Found «Не найден». Сервер не может найти запрашиваемый ресурс. Код этого
ответа, наверно, самый известный из-за частоты его появления в вебе. HTTP/0.9 и выше 405 Method Not Allowed «Метод не разрешён». Сервер знает о запрашиваемом методе, но он был
деактивирован и не может быть использован. Два обязательных метода,
GET и HEAD, никогда не должны быть
деактивированы и не должны возвращать этот код ошибки. Только HTTP/1.1 406 Not Acceptable

Этот ответ отсылается, когда веб сервер после выполнения
server-driven content negotiation, не нашёл контента, отвечающего критериям, полученным из user agent.

Только HTTP/1.1 407 Proxy Authentication Required Этот код ответа аналогичен коду 401, только аутентификация требуется для
прокси сервера. Только HTTP/1.1 408 Request Timeout Ответ с таким кодом может прийти, даже без предшествующего запроса. Он
означает, что сервер хотел бы отключить это неиспользуемое соединение.
Этот метод используется все чаще с тех пор, как некоторые браузеры,
вроде Chrome и IE9, стали использовать
HTTP механизмы предварительного соединения
для ускорения сёрфинга (смотрите баг 634278, будущей
реализации этого механизма в Firefox). Также учитывайте, что некоторые
серверы прерывают соединения не отправляя подобных сообщений. Только HTTP/1.1 409 Conflict

Этот ответ отсылается, когда запрос конфликтует с текущим состоянием
сервера.

Только HTTP/1.1 410 Gone

Этот ответ отсылается, когда запрашиваемый контент удалён с сервера.

Только HTTP/1.1 411 Length Required

Запрос отклонён, потому что сервер требует указание заголовка
Content-Length, но он не указан.

Только HTTP/1.1 412 Precondition Failed Клиент указал в своих заголовках условия, которые сервер не может
выполнить Только HTTP/1.1 413 Request Entity Too Large

Размер запроса превышает лимит, объявленный сервером. Сервер может
закрыть соединение, вернув заголовок Retry-After

Только HTTP/1.1 414 Request-URI Too Long URI запрашиваемый клиентом слишком длинный для того, чтобы сервер смог
его обработать Только HTTP/1.1 415 Unsupported Media Type Медиа формат запрашиваемых данных не поддерживается сервером, поэтому
запрос отклонён Только HTTP/1.1 416 Requested Range Not Satisfiable Диапазон указанный заголовком запроса Range не может быть
выполнен; возможно, он выходит за пределы переданного URI Только HTTP/1.1 417 Expectation Failed Этот код ответа означает, что ожидание, полученное из заголовка запроса
Expect, не может быть выполнено сервером. Только HTTP/1.1 Серверные 500 Internal Server Error «Внутренняя ошибка сервера». Сервер столкнулся с ситуацией, которую он
не знает как обработать. HTTP/0.9 и выше 501 Not Implemented «Не реализовано». Метод запроса не поддерживается сервером и не может быть
обработан. Единственные методы, которые сервера должны поддерживать (и,
соответственно, не должны возвращать этот код) — GET и
HEAD. HTTP/0.9 и выше 502 Bad Gateway «Плохой шлюз». Эта ошибка означает что сервер, во время работы в
качестве шлюза для получения ответа, нужного для обработки запроса,
получил недействительный (недопустимый) ответ. HTTP/0.9 и выше 503 Service Unavailable «Сервис недоступен». Сервер не готов обрабатывать запрос. Зачастую
причинами являются отключение сервера или то, что он перегружен.
Обратите внимание, что вместе с этим ответом удобная для
пользователей(user-friendly) страница должна отправлять объяснение
проблемы. Этот ответ должен использоваться для временных условий и
Retry-After: HTTP-заголовок должен, если возможно,
содержать предполагаемое время до восстановления сервиса. Веб-мастер
также должен позаботиться о заголовках, связанных с кешем, которые
отправляются вместе с этим ответом, так как эти ответы, связанные с
временными условиями, обычно не должны кешироваться. HTTP/0.9 и выше 504 Gateway Timeout Этот ответ об ошибке предоставляется, когда сервер действует как шлюз и
не может получить ответ вовремя. Только HTTP/1.1 505 HTTP Version Not Supported «HTTP-версия не поддерживается». HTTP-версия, используемая в запросе, не
поддерживается сервером. Только HTTP/1.1

From Wikipedia, the free encyclopedia

This is a list of Hypertext Transfer Protocol (HTTP) response status codes. Status codes are issued by a server in response to a client’s request made to the server. It includes codes from IETF Request for Comments (RFCs), other specifications, and some additional codes used in some common applications of the HTTP. The first digit of the status code specifies one of five standard classes of responses. The optional message phrases shown are typical, but any human-readable alternative may be provided, or none at all.

Unless otherwise stated, the status code is part of the HTTP standard (RFC 9110).

The Internet Assigned Numbers Authority (IANA) maintains the official registry of HTTP status codes.^[1]

All HTTP response status codes are separated into five classes or categories. The first digit of the status code defines the class of response, while the last two digits do not have any classifying or categorization role. There are five classes defined by the standard:

• 1xx informational response – the request was received, continuing process
• 2xx successful – the request was successfully received, understood, and accepted
• 3xx redirection – further action needs to be taken in order to complete the request
• 4xx client error – the request contains bad syntax or cannot be fulfilled
• 5xx server error – the server failed to fulfil an apparently valid request

1xx informational response

An informational response indicates that the request was received and understood. It is issued on a provisional basis while request processing continues. It alerts the client to wait for a final response. The message consists only of the status line and optional header fields, and is terminated by an empty line. As the HTTP/1.0 standard did not define any 1xx status codes, servers must not^{[note 1]} send a 1xx response to an HTTP/1.0 compliant client except under experimental conditions.

100 Continue

The server has received the request headers and the client should proceed to send the request body (in the case of a request for which a body needs to be sent; for example, a POST request). Sending a large request body to a server after a request has been rejected for inappropriate headers would be inefficient. To have a server check the request’s headers, a client must send Expect: 100-continue as a header in its initial request and receive a 100 Continue status code in response before sending the body. If the client receives an error code such as 403 (Forbidden) or 405 (Method Not Allowed) then it should not send the request’s body. The response 417 Expectation Failed indicates that the request should be repeated without the Expect header as it indicates that the server does not support expectations (this is the case, for example, of HTTP/1.0 servers).^[2]

101 Switching Protocols

The requester has asked the server to switch protocols and the server has agreed to do so.

102 Processing (WebDAV; RFC 2518)

A WebDAV request may contain many sub-requests involving file operations, requiring a long time to complete the request. This code indicates that the server has received and is processing the request, but no response is available yet.^[3] This prevents the client from timing out and assuming the request was lost.

103 Early Hints (RFC 8297)

Used to return some response headers before final HTTP message.^[4]

2xx success

This class of status codes indicates the action requested by the client was received, understood, and accepted.^[1]

200 OK

Standard response for successful HTTP requests. The actual response will depend on the request method used. In a GET request, the response will contain an entity corresponding to the requested resource. In a POST request, the response will contain an entity describing or containing the result of the action.

201 Created

The request has been fulfilled, resulting in the creation of a new resource.^[5]

202 Accepted

The request has been accepted for processing, but the processing has not been completed. The request might or might not be eventually acted upon, and may be disallowed when processing occurs.

203 Non-Authoritative Information (since HTTP/1.1)

The server is a transforming proxy (e.g. a Web accelerator) that received a 200 OK from its origin, but is returning a modified version of the origin’s response.^[6][7]

204 No Content

The server successfully processed the request, and is not returning any content.

205 Reset Content

The server successfully processed the request, asks that the requester reset its document view, and is not returning any content.

206 Partial Content

The server is delivering only part of the resource (byte serving) due to a range header sent by the client. The range header is used by HTTP clients to enable resuming of interrupted downloads, or split a download into multiple simultaneous streams.

207 Multi-Status (WebDAV; RFC 4918)

The message body that follows is by default an XML message and can contain a number of separate response codes, depending on how many sub-requests were made.^[8]

208 Already Reported (WebDAV; RFC 5842)

The members of a DAV binding have already been enumerated in a preceding part of the (multistatus) response, and are not being included again.

226 IM Used (RFC 3229)

The server has fulfilled a request for the resource, and the response is a representation of the result of one or more instance-manipulations applied to the current instance.^[9]

3xx redirection

This class of status code indicates the client must take additional action to complete the request. Many of these status codes are used in URL redirection.^[1]

A user agent may carry out the additional action with no user interaction only if the method used in the second request is GET or HEAD. A user agent may automatically redirect a request. A user agent should detect and intervene to prevent cyclical redirects.^[10]

300 Multiple Choices

Indicates multiple options for the resource from which the client may choose (via agent-driven content negotiation). For example, this code could be used to present multiple video format options, to list files with different filename extensions, or to suggest word-sense disambiguation.

301 Moved Permanently

This and all future requests should be directed to the given URI.

302 Found (Previously «Moved temporarily»)

Tells the client to look at (browse to) another URL. The HTTP/1.0 specification (RFC 1945) required the client to perform a temporary redirect with the same method (the original describing phrase was «Moved Temporarily»),^[11] but popular browsers implemented 302 redirects by changing the method to GET. Therefore, HTTP/1.1 added status codes 303 and 307 to distinguish between the two behaviours.^[10]

303 See Other (since HTTP/1.1)

The response to the request can be found under another URI using the GET method. When received in response to a POST (or PUT/DELETE), the client should presume that the server has received the data and should issue a new GET request to the given URI.

304 Not Modified

Indicates that the resource has not been modified since the version specified by the request headers If-Modified-Since or If-None-Match. In such case, there is no need to retransmit the resource since the client still has a previously-downloaded copy.

305 Use Proxy (since HTTP/1.1)

The requested resource is available only through a proxy, the address for which is provided in the response. For security reasons, many HTTP clients (such as Mozilla Firefox and Internet Explorer) do not obey this status code.

306 Switch Proxy

No longer used. Originally meant «Subsequent requests should use the specified proxy.»

307 Temporary Redirect (since HTTP/1.1)

In this case, the request should be repeated with another URI; however, future requests should still use the original URI. In contrast to how 302 was historically implemented, the request method is not allowed to be changed when reissuing the original request. For example, a POST request should be repeated using another POST request.

308 Permanent Redirect

This and all future requests should be directed to the given URI. 308 parallel the behaviour of 301, but does not allow the HTTP method to change. So, for example, submitting a form to a permanently redirected resource may continue smoothly.

4xx client errors

This class of status code is intended for situations in which the error seems to have been caused by the client. Except when responding to a HEAD request, the server should include an entity containing an explanation of the error situation, and whether it is a temporary or permanent condition. These status codes are applicable to any request method. User agents should display any included entity to the user.

400 Bad Request

The server cannot or will not process the request due to an apparent client error (e.g., malformed request syntax, size too large, invalid request message framing, or deceptive request routing).

401 Unauthorized

Similar to 403 Forbidden, but specifically for use when authentication is required and has failed or has not yet been provided. The response must include a WWW-Authenticate header field containing a challenge applicable to the requested resource. See Basic access authentication and Digest access authentication. 401 semantically means «unauthorised», the user does not have valid authentication credentials for the target resource.

Some sites incorrectly issue HTTP 401 when an IP address is banned from the website (usually the website domain) and that specific address is refused permission to access a website.^{[citation needed]}

402 Payment Required

Reserved for future use. The original intention was that this code might be used as part of some form of digital cash or micropayment scheme, as proposed, for example, by GNU Taler,^[13] but that has not yet happened, and this code is not widely used. Google Developers API uses this status if a particular developer has exceeded the daily limit on requests.^[14] Sipgate uses this code if an account does not have sufficient funds to start a call.^[15] Shopify uses this code when the store has not paid their fees and is temporarily disabled.^[16] Stripe uses this code for failed payments where parameters were correct, for example blocked fraudulent payments.^[17]

403 Forbidden

The request contained valid data and was understood by the server, but the server is refusing action. This may be due to the user not having the necessary permissions for a resource or needing an account of some sort, or attempting a prohibited action (e.g. creating a duplicate record where only one is allowed). This code is also typically used if the request provided authentication by answering the WWW-Authenticate header field challenge, but the server did not accept that authentication. The request should not be repeated.

404 Not Found

The requested resource could not be found but may be available in the future. Subsequent requests by the client are permissible.

405 Method Not Allowed

A request method is not supported for the requested resource; for example, a GET request on a form that requires data to be presented via POST, or a PUT request on a read-only resource.

406 Not Acceptable

The requested resource is capable of generating only content not acceptable according to the Accept headers sent in the request. See Content negotiation.

407 Proxy Authentication Required

The client must first authenticate itself with the proxy.

408 Request Timeout

The server timed out waiting for the request. According to HTTP specifications: «The client did not produce a request within the time that the server was prepared to wait. The client MAY repeat the request without modifications at any later time.»

409 Conflict

Indicates that the request could not be processed because of conflict in the current state of the resource, such as an edit conflict between multiple simultaneous updates.

410 Gone

Indicates that the resource requested was previously in use but is no longer available and will not be available again. This should be used when a resource has been intentionally removed and the resource should be purged. Upon receiving a 410 status code, the client should not request the resource in the future. Clients such as search engines should remove the resource from their indices. Most use cases do not require clients and search engines to purge the resource, and a «404 Not Found» may be used instead.

411 Length Required

The request did not specify the length of its content, which is required by the requested resource.

412 Precondition Failed

The server does not meet one of the preconditions that the requester put on the request header fields.

413 Payload Too Large

The request is larger than the server is willing or able to process. Previously called «Request Entity Too Large» in RFC 2616.^[18]

414 URI Too Long

The URI provided was too long for the server to process. Often the result of too much data being encoded as a query-string of a GET request, in which case it should be converted to a POST request. Called «Request-URI Too Long» previously in RFC 2616.^[19]

415 Unsupported Media Type

The request entity has a media type which the server or resource does not support. For example, the client uploads an image as image/svg+xml, but the server requires that images use a different format.

416 Range Not Satisfiable

The client has asked for a portion of the file (byte serving), but the server cannot supply that portion. For example, if the client asked for a part of the file that lies beyond the end of the file. Called «Requested Range Not Satisfiable» previously RFC 2616.^[20]

417 Expectation Failed

The server cannot meet the requirements of the Expect request-header field.^[21]

418 I’m a teapot (RFC 2324, RFC 7168)

This code was defined in 1998 as one of the traditional IETF April Fools’ jokes, in RFC 2324, Hyper Text Coffee Pot Control Protocol, and is not expected to be implemented by actual HTTP servers. The RFC specifies this code should be returned by teapots requested to brew coffee.^[22] This HTTP status is used as an Easter egg in some websites, such as Google.com’s «I’m a teapot» easter egg.^[23][24][25] Sometimes, this status code is also used as a response to a blocked request, instead of the more appropriate 403 Forbidden.^[26][27]

421 Misdirected Request

The request was directed at a server that is not able to produce a response (for example because of connection reuse).

422 Unprocessable Entity

The request was well-formed but was unable to be followed due to semantic errors.^[8]

423 Locked (WebDAV; RFC 4918)

The resource that is being accessed is locked.^[8]

424 Failed Dependency (WebDAV; RFC 4918)

The request failed because it depended on another request and that request failed (e.g., a PROPPATCH).^[8]

425 Too Early (RFC 8470)

Indicates that the server is unwilling to risk processing a request that might be replayed.

426 Upgrade Required

The client should switch to a different protocol such as TLS/1.3, given in the Upgrade header field.

428 Precondition Required (RFC 6585)

The origin server requires the request to be conditional. Intended to prevent the ‘lost update’ problem, where a client GETs a resource’s state, modifies it, and PUTs it back to the server, when meanwhile a third party has modified the state on the server, leading to a conflict.^[28]

429 Too Many Requests (RFC 6585)

The user has sent too many requests in a given amount of time. Intended for use with rate-limiting schemes.^[28]

431 Request Header Fields Too Large (RFC 6585)

The server is unwilling to process the request because either an individual header field, or all the header fields collectively, are too large.^[28]

451 Unavailable For Legal Reasons (RFC 7725)

A server operator has received a legal demand to deny access to a resource or to a set of resources that includes the requested resource.^[29] The code 451 was chosen as a reference to the novel Fahrenheit 451 (see the Acknowledgements in the RFC).

5xx server errors

The server failed to fulfil a request.

Response status codes beginning with the digit «5» indicate cases in which the server is aware that it has encountered an error or is otherwise incapable of performing the request. Except when responding to a HEAD request, the server should include an entity containing an explanation of the error situation, and indicate whether it is a temporary or permanent condition. Likewise, user agents should display any included entity to the user. These response codes are applicable to any request method.

500 Internal Server Error

A generic error message, given when an unexpected condition was encountered and no more specific message is suitable.

501 Not Implemented

The server either does not recognize the request method, or it lacks the ability to fulfil the request. Usually this implies future availability (e.g., a new feature of a web-service API).

502 Bad Gateway

The server was acting as a gateway or proxy and received an invalid response from the upstream server.

503 Service Unavailable

The server cannot handle the request (because it is overloaded or down for maintenance). Generally, this is a temporary state.^[30]

504 Gateway Timeout

The server was acting as a gateway or proxy and did not receive a timely response from the upstream server.

505 HTTP Version Not Supported

The server does not support the HTTP version used in the request.

506 Variant Also Negotiates (RFC 2295)

Transparent content negotiation for the request results in a circular reference.^[31]

507 Insufficient Storage (WebDAV; RFC 4918)

The server is unable to store the representation needed to complete the request.^[8]

508 Loop Detected (WebDAV; RFC 5842)

The server detected an infinite loop while processing the request (sent instead of 208 Already Reported).

510 Not Extended (RFC 2774)

Further extensions to the request are required for the server to fulfill it.^[32]

511 Network Authentication Required (RFC 6585)

The client needs to authenticate to gain network access. Intended for use by intercepting proxies used to control access to the network (e.g., «captive portals» used to require agreement to Terms of Service before granting full Internet access via a Wi-Fi hotspot).^[28]

Unofficial codes

The following codes are not specified by any standard.

419 Page Expired (Laravel Framework)

Used by the Laravel Framework when a CSRF Token is missing or expired.

420 Method Failure (Spring Framework)

A deprecated response used by the Spring Framework when a method has failed.^[33]

420 Enhance Your Calm (Twitter)

Returned by version 1 of the Twitter Search and Trends API when the client is being rate limited; versions 1.1 and later use the 429 Too Many Requests response code instead.^[34] The phrase «Enhance your calm» comes from the 1993 movie Demolition Man, and its association with this number is likely a reference to cannabis.^{[citation needed]}

430 Request Header Fields Too Large (Shopify)

Used by Shopify, instead of the 429 Too Many Requests response code, when too many URLs are requested within a certain time frame.^[35]

450 Blocked by Windows Parental Controls (Microsoft)

The Microsoft extension code indicated when Windows Parental Controls are turned on and are blocking access to the requested webpage.^[36]

498 Invalid Token (Esri)

Returned by ArcGIS for Server. Code 498 indicates an expired or otherwise invalid token.^[37]

499 Token Required (Esri)

Returned by ArcGIS for Server. Code 499 indicates that a token is required but was not submitted.^[37]

509 Bandwidth Limit Exceeded (Apache Web Server/cPanel)

The server has exceeded the bandwidth specified by the server administrator; this is often used by shared hosting providers to limit the bandwidth of customers.^[38]

529 Site is overloaded

Used by Qualys in the SSLLabs server testing API to signal that the site can’t process the request.^[39]

530 Site is frozen

Used by the Pantheon Systems web platform to indicate a site that has been frozen due to inactivity.^[40]

598 (Informal convention) Network read timeout error

Used by some HTTP proxies to signal a network read timeout behind the proxy to a client in front of the proxy.^[41]

599 Network Connect Timeout Error

An error used by some HTTP proxies to signal a network connect timeout behind the proxy to a client in front of the proxy.

Internet Information Services

Microsoft’s Internet Information Services (IIS) web server expands the 4xx error space to signal errors with the client’s request.

440 Login Time-out

The client’s session has expired and must log in again.^[42]

449 Retry With

The server cannot honour the request because the user has not provided the required information.^[43]

451 Redirect

Used in Exchange ActiveSync when either a more efficient server is available or the server cannot access the users’ mailbox.^[44] The client is expected to re-run the HTTP AutoDiscover operation to find a more appropriate server.^[45]

IIS sometimes uses additional decimal sub-codes for more specific information,^[46] however these sub-codes only appear in the response payload and in documentation, not in the place of an actual HTTP status code.

nginx

The nginx web server software expands the 4xx error space to signal issues with the client’s request.^[47][48]

444 No Response

Used internally^[49] to instruct the server to return no information to the client and close the connection immediately.

494 Request header too large

Client sent too large request or too long header line.

495 SSL Certificate Error

An expansion of the 400 Bad Request response code, used when the client has provided an invalid client certificate.

496 SSL Certificate Required

An expansion of the 400 Bad Request response code, used when a client certificate is required but not provided.

497 HTTP Request Sent to HTTPS Port

An expansion of the 400 Bad Request response code, used when the client has made a HTTP request to a port listening for HTTPS requests.

499 Client Closed Request

Used when the client has closed the request before the server could send a response.

Cloudflare

Cloudflare’s reverse proxy service expands the 5xx series of errors space to signal issues with the origin server.^[50]

520 Web Server Returned an Unknown Error

The origin server returned an empty, unknown, or unexpected response to Cloudflare.^[51]

521 Web Server Is Down

The origin server refused connections from Cloudflare. Security solutions at the origin may be blocking legitimate connections from certain Cloudflare IP addresses.

522 Connection Timed Out

Cloudflare timed out contacting the origin server.

523 Origin Is Unreachable

Cloudflare could not reach the origin server; for example, if the DNS records for the origin server are incorrect or missing.

524 A Timeout Occurred

Cloudflare was able to complete a TCP connection to the origin server, but did not receive a timely HTTP response.

525 SSL Handshake Failed

Cloudflare could not negotiate a SSL/TLS handshake with the origin server.

526 Invalid SSL Certificate

Cloudflare could not validate the SSL certificate on the origin web server. Also used by Cloud Foundry’s gorouter.

527 Railgun Error

Error 527 indicates an interrupted connection between Cloudflare and the origin server’s Railgun server.^[52]

530

Error 530 is returned along with a 1xxx error.^[53]

AWS Elastic Load Balancer

Amazon’s Elastic Load Balancing adds a few custom return codes

460

Client closed the connection with the load balancer before the idle timeout period elapsed. Typically when client timeout is sooner than the Elastic Load Balancer’s timeout.^[54]

463

The load balancer received an X-Forwarded-For request header with more than 30 IP addresses.^[54]

561 Unauthorized

An error around authentication returned by a server registered with a load balancer. You configured a listener rule to authenticate users, but the identity provider (IdP) returned an error code when authenticating the user.^[55]

Caching warning codes (obsoleted)

The following caching related warning codes were specified under RFC 7234. Unlike the other status codes above, these were not sent as the response status in the HTTP protocol, but as part of the «Warning» HTTP header.^[56][57]

Since this «Warning» header is often neither sent by servers nor acknowledged by clients, this header and its codes were obsoleted by the HTTP Working Group in 2022 with RFC 9111.^[58]

110 Response is Stale

The response provided by a cache is stale (the content’s age exceeds a maximum age set by a Cache-Control header or heuristically chosen lifetime).

111 Revalidation Failed

The cache was unable to validate the response, due to an inability to reach the origin server.

112 Disconnected Operation

The cache is intentionally disconnected from the rest of the network.

113 Heuristic Expiration

The cache heuristically chose a freshness lifetime greater than 24 hours and the response’s age is greater than 24 hours.

199 Miscellaneous Warning

Arbitrary, non-specific warning. The warning text may be logged or presented to the user.

214 Transformation Applied

Added by a proxy if it applies any transformation to the representation, such as changing the content encoding, media type or the like.

299 Miscellaneous Persistent Warning

Same as 199, but indicating a persistent warning.

See also

• Custom error pages
• List of FTP server return codes
• List of HTTP header fields
• List of SMTP server return codes
• Common Log Format

Explanatory notes

References

External links

• «RFC 9110: HTTP Semantics and Content, Section 15 «Status Codes»«.
• Hypertext Transfer Protocol (HTTP) Status Code Registry