Upstream#

The module is used to define groups of servers that can be referenced by the proxy_pass, fastcgi_pass, uwsgi_pass, scgi_pass, memcached_pass and grpc_pass directives.

Configuration Example#

upstream backend {
    zone backend 1m;
    server backend1.example.com       weight=5;
    server backend2.example.com:8080;
    server backend3.example.com       service=_example._tcp resolve;
    server unix:/tmp/backend3;

    server backup1.example.com:8080   backup;
    server backup2.example.com:8080   backup;
}

resolver 127.0.0.53 status_zone=resolver;

server {
    location / {
        proxy_pass http://backend;
    }
}

Directives#

bind_conn (PRO)#

Syntax

bind_conn value;

Default

Context

upstream

Enables binding the server connection to the client when the value, which is set as a string of variables, becomes anything other than "" and "0".

Attention

The bind_conn directive must be used after all directives that set the load balancing method; otherwise, it won't work. If sticky is also used, bind_conn should appear after sticky.

Attention

When using the directive, configure the http_proxy module to allow keepalive connections, for example:

proxy_http_version 1.1;
proxy_set_header Connection "";

A typical use case for the directive is proxying NTLM-authenticated connections, where the client should be bound to the server when the negotiation starts:

map $http_authorization   $ntlm {
    ~*^N(?:TLM|egotiate)  1;
}

upstream ntlm_backend {
    server 127.0.0.1:8080;
    bind_conn $ntlm;
}

server {
    # ...
    location / {
        proxy_pass http://ntlm_backend;
        proxy_http_version 1.1;
        proxy_set_header Connection "";
        # ...
    }
}

feedback (PRO)#

Added in version 1.6.0: PRO

Syntax

feedback variable [inverse] [factor=number] [account=condition_variable] [last_byte];

Default

Context

upstream

Enables a feedback-based load balancing mechanism for the upstream. It adjusts the load balancing decisions dynamically, multiplying each peer's weight by its average feedback value that is affected by the value of a variable over time and is subject to an optional condition.

The following parameters are accepted:

variable

The variable from which the feedback value is taken. It should represent a performance or health metric, and is intended to be supplied by the peer in header fields or otherwise.

The value is assessed at each response from the peer and factored into the rolling average according to inverse and factor settings.

inverse

If set, the feedback value is interpreted inversely, meaning lower values indicate better performance.

factor

The factor by which the feedback value is weighted when calculating the average. Valid values are integers between 0 and 99. By default — 90.

The average feedback is calculated using the exponential moving average formula.

The larger is the factor, the less is the average affected by new values; if the factor is set to 90, the result has 90% of the previous value and only 10% of the new value.

account

Specifies a condition variable that controls which responses should be included in the calculation. The average is updated with the feedback value only if the condition variable for the response isn't "" or "0".

Note

By default, responses from probes aren't included in the calculation; combining the $upstream_probe variable with account allows to include these responses or even exclude everything else.

last_byte

Allows processing feedback from the upstream server after the full response has been received, instead of just after the header.

Example:

upstream backend {

    zone backend 1m;

    feedback $feedback_value factor=80 account=$condition_value;

    server backend1.example.com;
    server backend2.example.com;
}

map $upstream_http_custom_score $feedback_value {
    "high"                      100;
    "medium"                    75;
    "low"                       50;
    default                     10;
}

map $upstream_probe $condition_value {
    "high_priority" "1";
    "low_priority"  "0";
    default         "1";
}

This categorizes server responses into different feedback levels based on specific scores obtained from response header fields, and also adds a condition mapped from $upstream_probe to account only for the responses from the high_priority probe or responses to regular client requests.

hash#

Syntax

hash key [consistent];

Default

Context

upstream

Specifies a load balancing method for a server group where the client-server mapping is based on the hashed key value. The key can contain text, variables, and their combinations. Note that adding or removing a server from the group may result in remapping most of the keys to different servers. The method is compatible with the Cache::Memcached Perl library.

If the consistent parameter is specified, the ketama consistent hashing method will be used instead. The method ensures that only a few keys will be remapped to different servers when a server is added to or removed from the group. This helps to achieve a higher cache hit ratio for caching servers. The method is compatible with the Cache::Memcached::Fast Perl library with the ketama_points parameter set to 160.

ip_hash#

Syntax

ip_hash;

Default

Context

upstream

Specifies that a group should use a load balancing method where requests are distributed between servers based on client IP addresses. The first three octets of the client IPv4 address, or the entire IPv6 address, are used as a hashing key. The method ensures that requests from the same client will always be passed to the same server except when this server is unavailable. In the latter case client requests will be passed to another server. Most probably, it will always be the same server as well.

If one of the servers needs to be temporarily removed, it should be marked with the down parameter in order to preserve the current hashing of client IP addresses.

upstream backend {
    ip_hash;

    server backend1.example.com;
    server backend2.example.com;
    server backend3.example.com down;
    server backend4.example.com;
}

keepalive#

Syntax

keepalive connections;

Default

Context

upstream

Activates the cache for connections to upstream servers.

The connections parameter sets the maximum number of idle keepalive connections to upstream servers that are preserved in the cache of each worker process. When this number is exceeded, the least recently used connections are closed.

Note

It should be particularly noted that the keepalive directive does not limit the total number of connections to upstream servers that an Angie worker process can open. The connections parameter should be set to a number small enough to let upstream servers process new incoming connections as well.

Attention

The keepalive directive must be used after all directives that set the load balancing method; otherwise, it won't work.

Example configuration of memcached upstream with keepalive connections:

upstream memcached_backend {
    server 127.0.0.1:11211;
    server 10.0.0.2:11211;

    keepalive 32;
}

server {
    #...

    location /memcached/ {
        set $memcached_key $uri;
        memcached_pass memcached_backend;
    }

}

For HTTP, the proxy_http_version directive should be set to "1.1" and the "Connection" header field should be cleared:

upstream http_backend {
    server 127.0.0.1:8080;

    keepalive 16;
}

server {
    #...

    location /http/ {
        proxy_pass http://http_backend;
        proxy_http_version 1.1;
        proxy_set_header Connection "";
    #    ...
    }
}

Note

Alternatively, HTTP/1.0 persistent connections can be used by passing the "Connection: Keep-Alive" header field to an upstream server, though this method is not recommended.

For FastCGI servers, it is required to set fastcgi_keep_conn for keepalive connections to work:

upstream fastcgi_backend {
    server 127.0.0.1:9000;

    keepalive 8;
}

server {
    #...

    location /fastcgi/ {
        fastcgi_pass fastcgi_backend;
        fastcgi_keep_conn on;
    #    ...
    }
}

Note

SCGI and uwsgi protocols do not have a notion of keepalive connections.

keepalive_requests#

Syntax

keepalive_requests number;

Default

keepalive_requests 1000;

Context

upstream

Sets the maximum number of requests that can be served through one keepalive connection. After the maximum number of requests is made, the connection is closed.

Closing connections periodically is necessary to free per-connection memory allocations. Therefore, using too high maximum number of requests could result in excessive memory usage and not recommended.

keepalive_time#

Syntax

keepalive_time time;

Default

keepalive_time 1h;

Context

upstream

Limits the maximum time during which requests can be processed through one keepalive connection. After this time is reached, the connection is closed following the subsequent request processing.

keepalive_timeout#

Syntax

keepalive_timeout time;

Default

keepalive_timeout 60s;

Context

upstream

Sets a timeout during which an idle keepalive connection to an upstream server will stay open.

least_conn#

Syntax

least_conn;

Default

Context

upstream

Specifies that a group should use a load balancing method where a request is passed to the server with the least number of active connections, taking into account weights of servers. If there are several such servers, they are tried in turn using a weighted round-robin balancing method.

least_time (PRO)#

Syntax

least_time header | last_byte [factor=number] [account=condition_variable];

Default

Context

upstream

Specifies that the group should use a load balancing method where an active server's chance of receiving the request is inversely proportional to its average response time; the less it is, the more requests a server gets.

header

The directive accounts for response headers only.

last_byte

The directive uses the average time to receive the entire response.

Added in version 1.7.0: PRO

factor

Serves the same purpose as response_time_factor (PRO) and overrides it if set.

account

Specifies a condition variable that controls which responses should be included in the calculation. The average is updated only if the condition variable for the response isn't "" or "0".

Note

By default, responses from probes aren't included in the calculation; combining the $upstream_probe variable with account allows to include these responses or even exclude everything else.

The respective moving averages, adjusted for factor and account, are also presented as header_time and response_time in the server's health object among the upstream metrics in the API.

queue (PRO)#

Added in version 1.4.0: PRO

Syntax

queue number [timeout=time];

Default

Context

upstream

If it is not possible to assign a proxied server to a request on the first attempt (for example, during a brief service interruption or when there is a surge in load reaching the max_conns limit), the request is not rejected; instead, Angie attempts to enqueue it for processing.

The number in the directive sets the maximum number of requests in the queue for a worker process. If the queue is full, a 502 (Bad Gateway) error is returned to the client.

Note

The logic of the proxy_next_upstream directive also applies to queued requests. Specifically, if a server was selected for a request but it cannot be handed over to it, the request may be returned to the queue.

If a server is not selected to process a queued request within the time set by timeout (default is 60 seconds), a 502 (Bad Gateway) error is returned to the client. Requests from clients that prematurely close the connection are also removed from the queue; there are counters for requests passing through the queue in the API.

Attention

The queue directive must be used after all directives that set the load balancing method; otherwise, it won't work.

random#

Syntax

random [two];

Default

Context

upstream

Specifies that a group should use a load balancing method where a request is passed to a randomly selected server, taking into account weights of servers.

The optional two parameter instructs Angie to randomly select two servers and then choose a server using the specified method. The default method is least_conn which passes a request to a server with the least number of active connections.

response_time_factor (PRO)#

Syntax

response_time_factor number;

Default

response_time_factor 90;

Context

upstream

If the least_time (PRO) load balancing method is used, sets the smoothing factor for the previous value when average response time is calculated using the exponential moving average formula.

The larger is the number, the less is the average affected by new values; if the number is set to 90, the result has 90% of the previous value and only 10% of the new value. The allowed range is 0 to 99, inclusive.

The respective moving averages are presented as header_time (headers only) and response_time (entire responses) in the server's health object among the upstream metrics in the API.

Note

The calculation accounts for successful reponses only; what is considered an unsuccessful response is defined by the proxy_next_upstream, fastcgi_next_upstream, uwsgi_next_upstream, scgi_next_upstream, memcached_next_upstream, and grpc_next_upstream directives. Besides, header_time is updated only if all headers are received and processed, and response_time is updated only if the entire reponse is received.

server#

Syntax

server address [parameters];

Default

Context

upstream

Defines the address and other parameters of a server. The address can be specified as a domain name or IP address, with an optional port, or as a UNIX domain socket path specified after the unix: prefix. If a port is not specified, the port 80 is used. A domain name that resolves to several IP addresses defines multiple servers at once.

The following parameters can be defined:

weight=number

sets the weight of the server
by default, 1.

max_conns=number

limits the maximum number of simultaneous active connections to the proxied server.
Default value is 0, meaning there is no limit. If the server group does not reside in the shared memory, the limitation works per each worker process.

Note

If idle keepalive connections, multiple workers, and the shared memory are enabled, the total number of active and idle connections to the proxied server may exceed the max_conns value.

max_fails=number — sets the number of unsuccessful attempts to communicate with the server that should happen in the duration set by fail_timeout to consider the server unavailable; it is then retried after the same duration.

What is considered an unsuccessful attempt is defined by the proxy_next_upstream, fastcgi_next_upstream, uwsgi_next_upstream, scgi_next_upstream, memcached_next_upstream, and grpc_next_upstream directives.

When max_fails is reached, the peer is also considered unhealthy by the upstream_probe (PRO) probes; it won't receive client requests until the probes consider it healthy again.

Note

If a server in an upstream resolves into multiple peers, its max_fails setting applies to each peer individually.

If an upstream contains only one peer after all its server directives are resolved, the max_fails setting has no effect and will be ignored.

max_fails=1

the default number of unsuccessful attempts

max_fails=0

disables the accounting of attempts

fail_timeout=time — sets the period of time during which a number of unsuccessful attempts to communicate with the server (max_fails) should happen to consider the server unavailable. The server then becomes unavailable for the same amount of time before it is retried.

By default, this is set to 10 seconds.

Note

If a server in an upstream resolves into multiple peers, its fail_timeout setting applies to each peer individually.

If an upstream contains only one peer after all its server directives are resolved, the fail_timeout setting has no effect and will be ignored.

backup

marks the server as a backup server. It will be passed requests when the primary servers are unavailable.

down

marks the server as permanently unavailable.

drain

sets the server to draining; this means it receives only requests from the sessions that were bound earlier with sticky. Otherwise it behaves similarly to down.

Caution

The parameter backup cannot be used along with the hash, ip_hash, and random load balancing methods.

The down and drain options are mutually exclusive.

Added in version 1.1.0.

resolve

enables monitoring changes to the list of IP addresses that corresponds to a domain name, updating it without a configuration reload. The group should be stored in a shared memory zone; also, you need to define a resolver.

service=name

enables resolving DNS SRV records and sets the service name. For this parameter to work, specify the resolve server parameter, providing a hostname without a port number.

If there are no dots in the service name, the name is formed according to the RFC standard: the service name is prefixed with _, then _tcp is added after a dot. Thus, the service name http will result in _http._tcp.

Angie resolves the SRV records by combining the normalized service name and the hostname and obtaining the list of servers for the combination via DNS, along with their priorities and weights.

  • Top-priority SRV records (ones that share the minimum priority value) resolve into primary servers, and other records become backup servers. If backup is set with server, top-priority SRV records resolve into backup servers, and other records are ignored.

  • Weight influences the selection of servers by the assigned capacity: higher weights receive more requests. If set by both the server directive and the SRV record, the weight set by server is used.

This example will look up the _http._tcp.backend.example.com record:

server backend.example.com service=http resolve;

Added in version 1.2.0: Angie

Added in version 1.1.0-P1: Angie PRO

sid=id

sets the server ID within the group.
If the parameter is omitted, the ID is set to the hexadecimal MD5 hash value of either the IP address and port or the UNIX domain socket path.

Added in version 1.4.0.

slow_start=time

sets the time to recover the weight for a server that goes back online, if load balancing uses the round-robin or least_conn method.

If the value is set and the server is again considered available and healthy as defined by max_fails and upstream_probe (PRO), the server will steadily recover its designated weight within the allocated timeframe.

If the value isn't set, the server in a similar situation will recover its designated weight immediately.

Note

If there's only one server in an upstream, slow_start has no effect and will be ignored.

state (PRO)#

Added in version 1.2.0: PRO

Syntax

state file;

Default

Context

upstream

Specifies the file where the upstream's server list is persisted. When installing from our packages, a designated /var/lib/angie/state/ (/var/db/angie/state/ on FreeBSD) directory with appropriate permissions is created to store these files, so you will only need to add the file's basename in the configuration:

upstream backend {

    zone backend 1m;
    state /var/lib/angie/state/<FILE NAME>;
}

The format of this server list is similar to server. The contents of the file change whenever there is any modification to servers in the /config/http/upstreams/ section via the configuration API. The file is read at Angie start or configuration reload.

Caution

For the state directive to be used in an upstream block, the block should have no server directives; instead, it must have a shared memory zone (zone).

sticky#

Added in version 1.2.0: Angie

Added in version 1.1.0-P1: Angie PRO

Syntax

sticky cookie name [attr=value]...;

sticky route $variable...;

sticky learn zone=zone create=$create_var1... lookup=$lookup_var1... [header] [timeout=time];

Default

Context

upstream

Configures the binding of client sessions to proxied servers in the mode specified by the first parameter; to drain requests from servers that have sticky defined, use the drain option in the server block.

Attention

The sticky directive must be used after all directives that set the load balancing method; otherwise, it won't work. If bind_conn (PRO) is also used, bind_conn should appear after sticky.

This mode uses cookies to maintain session persistence. It is more suitable for situations where cookies are already used for session management.

Here, a client's request, not yet bound to any server, is sent to a server chosen according to the configured balancing method. Also, Angie sets a cookie with a unique value identifying the server.

The cookie's name (name) is set by the sticky directive, and the value (value) corresponds to the sid parameter of the server directive. Note that the parameter is additionally hashed if the sticky_secret directive is set.

Subsequent client requests that contain this cookie are forwarded to the server identified by the cookie's value, which is the server with the specified sid. If selecting a server fails or the chosen server can't handle the request, another server is selected according to the configured balancing method.

The directive allows assigning attributes to the cookie; the only attribute set by default is path=/. Attribute values are specified as strings with variables. To remove an attribute, set an empty value for it: attr=. Thus, sticky cookie path= creates a cookie without path.

Here, Angie creates a cookie named srv_id with a one-hour lifespan and a variable-specified domain:

upstream backend {
    server backend1.example.com:8080;
    server backend2.example.com:8080;

    sticky cookie srv_id domain=$my_domain max-age=3600;
}

sticky_secret#

Added in version 1.2.0: Angie

Added in version 1.1.0-P1: Angie PRO

Syntax

sticky_secret string;

Default

Context

upstream

Adds the string as the salt value to the MD5 hashing function for the sticky directive in cookie and route modes. The string may contain variables, for example, $remote_addr:

upstream backend {
    server backend1.example.com:8080;
    server backend2.example.com:8080;

    sticky cookie cookie_name;
    sticky_secret my_secret.$remote_addr;
}

Salt is appended to the value being hashed; to verify the hashing mechanism independently:

$ echo -n "<VALUE><SALT>" | md5sum

sticky_strict#

Added in version 1.2.0: Angie

Added in version 1.1.0-P1: Angie PRO

Syntax

sticky_strict on | off;

Default

sticky_strict off;

Context

upstream

When enabled, causes Angie to return an HTTP 502 error to the client if the desired server is unavailable, rather than using any other available server as it would when no servers in the upstream are available.

upstream#

Syntax

upstream name { ... }

Default

Context

http

Defines a group of servers. Servers can listen on different ports. In addition, servers listening on TCP and UNIX domain sockets can be mixed.

Example:

upstream backend {
    server backend1.example.com weight=5;
    server 127.0.0.1:8080       max_fails=3 fail_timeout=30s;
    server unix:/tmp/backend3;

    server backup1.example.com  backup;
}

By default, requests are distributed between the servers using a weighted round-robin balancing method. In the above example, each 7 requests will be distributed as follows: 5 requests go to backend1.example.com and one request to each of the second and third servers.

If an error occurs during communication with a server, the request will be passed to the next server, and so on until all of the functioning servers will be tried. If a successful response could not be obtained from any of the servers, the client will receive the result of the communication with the last server.

zone#

Syntax

zone name [size];

Default

Context

upstream

Defines the name and size of the shared memory zone that keeps the group's configuration and run-time state that are shared between worker processes. Several groups may share the same zone. In this case, it is enough to specify the size only once.

Built-in Variables#

The http_upstream module supports the following built-in variables:

$upstream_addr#

keeps the IP address and port, or the path to the UNIX domain socket of the upstream server. If several servers were contacted during request processing, their addresses are separated by commas, e.g. :

192.168.1.1:80, 192.168.1.2:80, unix:/tmp/sock

If an internal redirect from one server group to another happens, initiated by "X-Accel-Redirect" or error_page, then the server addresses from different groups are separated by colons, e.g.:

192.168.1.1:80, 192.168.1.2:80, unix:/tmp/sock : 192.168.10.1:80, 192.168.10.2:80

If a server cannot be selected, the variable keeps the name of the server group.

$upstream_bytes_received#

number of bytes received from an upstream server. Values from several connections are separated by commas and colons like addresses in the $upstream_addr variable.

$upstream_bytes_sent#

number of bytes sent to an upstream server. Values from several connections are separated by commas and colons like addresses in the $upstream_addr variable.

$upstream_cache_status#

keeps the status of accessing a response cache. The status can be either MISS, BYPASS, EXPIRED, STALE, UPDATING, REVALIDATED or HIT:

  • MISS: The response isn't found in the cache, and the request is forwarded to the upstream server.

  • BYPASS: The cache is bypassed, and the request is directly forwarded to the upstream server.

  • EXPIRED: The cached response is stale, and a new request for the updated content is sent to the upstream server.

  • STALE: The cached response is stale, but will be served to the clients until an update has been eventually fetched from the upstream server.

  • UPDATING: The cached response is stale, but will be served to the clients until the currently ongoing update from the upstream server has been finished.

  • REVALIDATED: The cached response is stale, but is successfully revalidated and doesn't need an update from the upstream server.

  • HIT: The response was served from the cache.

If the cache was bypassed entirely without accessing it, the variable isn't set.

$upstream_connect_time#

keeps time spent on establishing a connection with the upstream server; the time is kept in seconds with millisecond resolution. In case of SSL, includes time spent on handshake. Times of several connections are separated by commas and colons like addresses in the $upstream_addr variable.

$upstream_header_time#

stores time spent on receiving the response header from the upstream server; the time is kept in seconds with millisecond resolution. Times of several responses are separated by commas and colons like addresses in the $upstream_addr variable.

$upstream_http_<name>#

stores server response header fields. For example, the "Server" response header field is available through the $upstream_http_server variable. The rules of converting header field names to variable names are the same as for the variables that start with the "$http_" prefix. Only the header fields from the response of the last server are saved.

$upstream_queue_time#

stores time the request spent in the queue before a server was selected; the time is kept in seconds with millisecond resolution. Times of several selection attempts are separated by commas and colons, like addresses in the $upstream_addr variable.

$upstream_response_length#

keeps the length of the response obtained from the upstream server; the length is kept in bytes. Lengths of several responses are separated by commas and colons like addresses in the $upstream_addr variable.

$upstream_response_time#

keeps time spent on receiving the response from the upstream server; the time is kept in seconds with millisecond resolution. Times of several responses are separated by commas and colons like addresses in the $upstream_addr variable.

$upstream_status#

keeps status code of the response obtained from the upstream server. Status codes of several responses are separated by commas and colons like addresses in the $upstream_addr variable. If a server cannot be selected, the variable keeps the 502 (Bad Gateway) status code.

$upstream_sticky_status#

Status of sticky requests.

""

Request sent to upstream without sticky enabled.

NEW

Request without sticky information.

HIT

Request with sticky information routed to the desired backend.

MISS

Request with sticky information routed to the backend selected by the load balancing algorithm.

Values from multiple connections are separated by commas and colons, similar to addresses in the $upstream_addr variable.

$upstream_trailer_<name>#

stores fields from the end of the response obtained from the upstream server.