proxy-server.conf(5)

📦
version 2.38.2 · 4 versions
, in debian trixie, nixos 26.05, nixpkgs unstable, ubuntu questing, ubuntu resolute, ubuntu stonking

💖 proxy-server.conf(5) 💖

the OpenStack Swift proxy server configuration file — yes bestie, this is the one, the manual page, section 5, and we're about to read the whole thing together 🫶

✨ SYNOPSIS ✨

proxy-server.conf

🌸 DESCRIPTION 🌸

Okay so. This is the config file that the proxy server uses, and also the one that all the other proxy middlewares use. She's popular. She's booked and busy. 👑

The file follows the python-pastedeploy syntax. That means it's chopped up into sections, and each section is wrapped in square brackets [like this]. Inside each section you get a certain number of key/value parameters, and we're gonna go through all of them later in this page, no skipping, we do not skip here. 💅

Any line that starts with a # symbol? Ignored. Gone. Not invited. 🚫

If you want more tea on the python-pastedeploy config format, it's literally right here:

https://docs.pylonsproject.org/projects/pastedeploy/en/latest/#config-format

👑 GLOBAL SECTION 👑

This one is signalled by the section name [DEFAULT]. Here's everything that's allowed to live in there, bestie, one by one:

  • bind_ip — the IP address the proxy server should bind to. Default is 0.0.0.0, which means it binds to literally every available address. she's everywhere. 🌈
  • bind_port — the TCP port the proxy server should bind to. Default is 80.
  • keep_idle — the value to set for socket TCP_KEEPIDLE. Default value is 600.
  • bind_timeout — timeout for binding the socket. Default is 30.
  • backlog — TCP backlog. That's the maximum number of allowed pending connections. Default value is 4096.
  • admin_key — the key used for admin calls that are HMAC signed. Default is empty, and empty means admin calls to /info get disabled. 🚫
  • disallowed_sections — this one lets you withhold sections so they don't show up in the public calls to /info. You can withhold subsections too by separating the dict level with a ".". The following would cause the sections 'container_quotas' and 'tempurl' to not be listed, and the key max_failed_deletes would be removed from bulk_delete. Default value is 'swift.valid_api_versions', which allows all registered features to be listed via HTTP GET /info — except the swift.valid_api_versions information. sneaky lil exception, love that for us. 🙈
  • workers — the number of pre-forked processes that will accept connections. Zero means no fork. The default is auto, which makes the server try to match the number of effective cpu cores if python multiprocessing is available (it's included with most python distributions >= 2.6), or fall back to one. And it's worth noting — fr fr — that individual workers will use many eventlet co-routines to service multiple concurrent requests. multitasking icon. ✨
  • max_clients — the maximum number of clients one worker can process simultaneously (it will actually accept(2) N + 1). Setting this to one (1) will only handle one request at a time, without accepting another request concurrently. The default is 1024.
  • user — the system user that the proxy server will run as. The default is swift.
  • expose_info — enables exposing configuration settings via HTTP GET /info. The default is true.
  • swift_dir — the Swift configuration directory. The default is /etc/swift.
  • cert_file — location of the SSL certificate file. Default path is /etc/swift/proxy.crt. This is disabled by default.
  • key_file — location of the SSL certificate key file. Default path is /etc/swift/proxy.key. This is disabled by default.
  • log_name — the label used when logging. The default is swift.
  • log_facility — syslog log facility. The default is LOG_LOCAL0.
  • log_level — logging level. The default is INFO.
  • log_address — logging address. The default is /dev/log.
  • log_max_line_length — caps the length of log lines to whatever value you give it. No limit if set to 0, which is the default.
  • log_headers — the default is false.
  • log_custom_handlers — a comma separated list of functions to call to set up custom log handlers. The functions get passed: conf, name, log_to_console, log_route, fmt, logger, adapted_logger. The default is empty.
  • log_udp_host — if it's set, log_udp_host will override log_address. 💥
  • log_udp_port — UDP log port, the default is 514.
  • log_statsd_host — the StatsD server. IPv4/IPv6 addresses and hostnames are both supported. If a hostname resolves to both an IPv4 and an IPv6 address, the IPv4 address will be used.
  • log_statsd_port — the default is 8125.
  • log_statsd_default_sample_rate — the default is 1.
  • log_statsd_sample_rate_factor — the default is 1.
  • log_statsd_metric_prefix — the default is empty.
  • client_timeout — the time to wait while receiving each chunk of data from a client or another backend node. The default is 60.
  • eventlet_debug — debug mode for the eventlet library. The default is false.
  • trans_id_suffix — this optional suffix (default is empty) gets appended to the swift transaction id, and it lets you easily figure out which cluster that X-Trans-Id belongs to. This is very useful when you're managing more than one swift cluster. chef's kiss for multi-cluster girlies. 💋
  • cors_allow_origin — a list of origin hosts that are allowed for CORS requests, in addition to whatever the container has set. Use a comma separated list of full URL (http://foo.bar:1234,https://foo.bar).
  • strict_cors_mode — if True (which is the default) then CORS requests are only allowed if their Origin header matches an allowed origin. Otherwise, any Origin is allowed. pick your lane bestie. 🎯
  • cors_expose_headers — a comma separated list of headers to expose through Access-Control-Expose-Headers, in addition to the defaults and any headers set in container metadata.
  • nice_priority — modifies the scheduling priority of server processes. Niceness values range from -20 (most favorable to the process) to 19 (least favorable to the process). The default does not modify priority.
  • ionice_class — modifies the I/O scheduling class of server processes. I/O niceness class values are IOPRIO_CLASS_RT (realtime), IOPRIO_CLASS_BE (best-effort) and IOPRIO_CLASS_IDLE (idle). The default does not modify class and priority. Work only with ionice_priority.
  • ionice_priority — modifies the I/O scheduling priority of server processes. I/O niceness priority is a number which goes from 0 to 7. The higher the value, the lower the I/O priority of the process. Work only with ionice_class. Ignored if IOPRIO_CLASS_IDLE is set.

🦋 PIPELINE SECTION 🦋

This one is signalled by the section name [pipeline:main]. Here's what's allowed in there:

  • pipeline — you use this when you need to apply a number of filters. It's a list of filters, ended by an application. The normal pipeline is:
"catch_errors gatekeeper healthcheck proxy-logging cache container_sync bulk tempurl ratelimit tempauth container-quotas account-quotas slo dlo versioned_writes proxy-logging proxy-server"
Note: the double proxy-logging in the pipeline is not a mistake. The left-most proxy-logging is there to log requests that were handled in middleware and never made it through to the right-most middleware (and proxy server). Double logging is prevented for normal requests. See proxy-logging docs. 💅

🎀 FILTER SECTION 🎀

Any section whose name is prefixed by "filter:" is a filter section. Filters are used to specify configuration parameters for specific swift middlewares. Below are the filters available and their respective acceptable parameters, and yes we are doing all of them. 🌸

[filter:healthcheck]

  • use — entry point for paste.deploy for the healthcheck middleware. This is the reference to the installed python egg. This is normally egg:swift#healthcheck.
  • disable_path — an optional filesystem path which, if present, will cause the healthcheck URL to return "503 Service Unavailable" with a body of "DISABLED BY FILE". she said no 💀

[filter:tempauth]

  • use — entry point for paste.deploy for the tempauth middleware. This is the reference to the installed python egg. This is normally egg:swift#tempauth.
  • set log_name — label used when logging. The default is tempauth.
  • set log_facility — syslog log facility. The default is LOG_LOCAL0.
  • set log_level — logging level. The default is INFO.
  • set log_address — logging address. The default is /dev/log.
  • set log_headers — enables the ability to log request headers. The default is False.
  • reseller_prefix — the reseller prefix will verify a token begins with this prefix before even attempting to validate it. Also, with authorization, only Swift storage accounts with this prefix will be authorized by this middleware. Useful if multiple auth systems are in use for one Swift cluster. The default is AUTH.
  • auth_prefix — the auth prefix will cause requests beginning with this prefix to be routed to the auth subsystem, for granting tokens, etc. The default is /auth/.
  • require_group — the require_group parameter names a group that must be presented by either X-Auth-Token or X-Service-Token. Usually this parameter is used only with multiple reseller prefixes (e.g., SERVICE_require_group=blah). By default, no group is needed.
Do not use .admin. 🚫
  • token_life — this is the time in seconds before the token expires. The default is 86400.
  • allow_overrides — this allows middleware higher in the WSGI pipeline to override auth processing, useful for middleware such as tempurl and formpost. If you know you're not going to use such middleware and you want a bit of extra security, you can set this to false. The default is true.
  • storage_url_scheme — this specifies what scheme to return with storage urls: http, https, or default (chooses based on what the server is running as). This can be useful with an SSL load balancer in front of a non-SSL server. balance that load bestie ⚖️
  • user_<account>_<user> — lastly, you need to list all the accounts/users you want here. The format is:
user_<account>_<user> = <key> [group] [group] [...] [storage_url]

or if you want underscores in <account> or <user>, you can base64 encode them (with no equal signs) and use this format:

user64_<account_b64>_<user_b64> = <key> [group] [group] [...] [storage_url]

okay that's the end of part 1, but we are NOT done, there's more coming so stay tuned bestie 💜

Okay bestie, so there are these special little groups 👑: .reseller_admin — who can do literally anything to any account for this auth — and .admin, who can do anything but only within the account. Two tiers of power, very main character vs. supporting cast 💅

And if neither of those groups is specified? Then the user can only touch containers that a .admin or a .reseller_admin has explicitly allowed for them. No sneaky access, periodt 🚫✨

That trailing optional storage_url? It lets you hand back an alternate URL to the user once they authenticate. If you don't specify it, it defaults to:

http[s]://<ip>:<port>/v1/<reseller_prefix>_<account>

…where http or https depends on whether cert_file is set in the [DEFAULT] section, <ip> and <port> come from the [DEFAULT] section's bind_ip and bind_port (and if those aren't there, they fall back to 127.0.0.1 and 8080 🌈), <reseller_prefix> is from this section, and <account> comes from the user_<account>_<user> name. iykyk 🧚

Here are the example entries, which are required for running the tests:

user_admin_admin = admin .admin .reseller_admin
user_test_tester = testing .admin
user_test2_tester2 = testing2 .admin
user_test_tester3 = testing3

✨ [filter:authtoken] ✨

To turn on Keystone authentication, you need to have the auth token middleware configured first — order matters, bestie 🎀. Here's an example below, and for all the different settings please go check Keystone's own documentation, okurrr?

You'll also need the keystoneauth middleware enabled and sitting in your main pipeline, so instead of having tempauth chilling in there, you can swap it to: authtoken keystoneauth 🦋

The auth credentials ("project_domain_name", "user_domain_name", "username", "project_name", "password") must match the Keystone credentials for the Swift service. The example values shown here assume a user named "swift" with admin role on a project named "service", both of them living in the Keystone domain with id "default". For other examples, refer to the KeystoneMiddleware documentation at https://docs.openstack.org/keystonemiddleware/latest/middlewarearchitecture.html#configuration 💜
paste.filter_factory =
keystonemiddleware.auth_token:filter_factory
www_authenticate_uri = http://keystonehost:5000
auth_url = http://keystonehost:5000
auth_plugin = password
project_domain_id = default
user_domain_id = default
project_name = service
username = swift
password = password

# delay_auth_decision defaults to False, but leaving it as
false will
# prevent other auth systems, staticweb, tempurl, formpost,
and ACLs from
# working. This value must be explicitly set to True.
delay_auth_decision = False
cache = swift.cache
include_service_catalog = False

✨ [filter:keystoneauth] ✨

Keystone authentication middleware. Your friendly neighborhood gatekeeper 🔥

  • use — Entry point for paste.deploy for the keystoneauth middleware. This is the reference to the installed python egg. This is normally egg:swift#keystoneauth.
  • reseller_prefix — This option lists account namespaces that this middleware is responsible for. The prefix gets placed before the Keystone project id. So for project 12345678 with prefix AUTH, the account is named AUTH_12345678 (i.e., the path is /v1/AUTH_12345678/...). You can have several prefixes by giving a comma-separated list, like: "reseller_prefix = AUTH, SERVICE". The empty string indicates a single blank/empty prefix. And if you need an empty prefix inside a list of prefixes, a value of '' (two single quote characters) indicates a blank/empty prefix. Except for that blank/empty prefix, an underscore ('_') character is appended to the value unless it's already there. 🎀
  • operator_roles — The user must have at least one role named by operator_roles on a project in order to create, delete and modify containers and objects, and to set and read privileged headers such as ACLs. If there are several reseller prefix items, you can prefix the parameter so it applies only to those accounts — for example the parameter SERVICE_operator_roles applies to the /v1/SERVICE_<project> path. If you omit the prefix, the option applies to all reseller prefix items. For the blank/empty prefix, prefix with '' (and do not put an underscore after the two single quote characters) 👑
  • reseller_admin_role — The reseller admin role has the ability to create and delete accounts. Big power, big responsibility 💅
  • allow_overrides — This allows middleware higher in the WSGI pipeline to override auth processing, which is useful for middleware such as tempurl and formpost. If you know you're not going to use such middleware and you want a bit of extra security, you can set this to false 🔒
  • service_roles — If the service_roles parameter is present, an X-Service-Token must be present in the request that, when validated, grants at least one role listed in the parameter. The X-Service-Token may be scoped to any project. If there are several reseller prefix items, you can prefix the parameter so it applies only to those accounts — for example the parameter SERVICE_service_roles applies to the /v1/SERVICE_<project> path. If you omit the prefix, the option applies to all reseller prefix items. For the blank/empty prefix, prefix with '' (do not put underscore after the two single quote characters). By default, no service_roles are required. ✨
  • default_domain_id — For backwards compatibility, keystoneauth will match names in cross-tenant access control lists (ACLs) when both the requesting user and the tenant are in the default domain, i.e. the domain to which existing tenants are migrated. The default_domain_id value configured here should be the same as the value used during migration of tenants to keystone domains. 🧚
  • allow_names_in_acls — For a new installation, or an installation in which keystone projects may move between domains, you should disable backwards compatible name matching in ACLs by setting allow_names_in_acls to false:

✨ [filter:cache] ✨

Caching middleware that manages caching in swift. It's the vibe-keeper of your data 🌸

  • use — Entry point for paste.deploy for the memcache middleware. This is the reference to the installed python egg. This is normally egg:swift#memcache.
  • set log_name — Label used when logging. The default is memcache.
  • set log_facility — Syslog log facility. The default is LOG_LOCAL0.
  • set log_level — Logging level. The default is INFO.
  • set log_address — Logging address. The default is /dev/log.
  • set log_headers — Enables the ability to log request headers. The default is False.
  • memcache_max_connections — Sets the maximum number of connections to each memcached server per worker.
  • memcache_servers — If not set in the configuration file, the value for memcache_servers will be read from /etc/swift/memcache.conf (see memcache.conf-sample) or, lacking that file, it will default to 127.0.0.1:11211. You can specify multiple servers separated with commas, as in: 10.1.2.3:11211,10.1.2.4:11211. (IPv6 addresses must follow rfc3986 section-3.2.2, i.e. [::1]:11211) 🌈

✨ [filter:ratelimit] ✨

Rate limits requests on both an Account and Container level. The limits are configurable, bestie — you're in charge 🎀

  • use — Entry point for paste.deploy for the ratelimit middleware. This is the reference to the installed python egg. This is normally egg:swift#ratelimit.
  • set log_name — Label used when logging. The default is ratelimit.
  • set log_facility — Syslog log facility. The default is LOG_LOCAL0.
  • set log_level — Logging level. The default is INFO.
  • set log_address — Logging address. The default is /dev/log.
  • set log_headers — Enables the ability to log request headers. The default is False.
  • clock_accuracy — This should represent how accurate the proxy servers' system clocks are with each other. 1000 means that all the proxies' clocks are accurate to each other within 1 millisecond. No ratelimit should be higher than the clock accuracy. The default is 1000.
  • max_sleep_time_seconds — App will immediately return a 498 response if the necessary sleep time ever exceeds the given max_sleep_time_seconds. The default is 60 seconds.
  • log_sleep_time_seconds — To allow visibility into rate limiting, set this value > 0 and all sleeps greater than the number will be logged. If set to 0 means disabled. The default is 0.
  • rate_buffer_seconds — Number of seconds the rate counter can drop and be allowed to catch up (at a faster than listed rate). A larger number will result in larger spikes in rate but better average accuracy. The default is 5.
  • account_ratelimit — If set, will limit PUT and DELETE requests to /account_name/container_name. Number is in requests per second. If set to 0 means disabled. The default is 0.
  • container_ratelimit_size — When set with container_limit_x = r: for containers of size x, limit requests per second to r. Will limit PUT, DELETE, and POST requests to /a/c/o. The default is ''.
  • container_listing_ratelimit_size — Similarly to the above container-level write limits, the following will limit container GET (listing) requests. 📋

✨ [filter:domain_remap] ✨

Middleware that translates container and account parts of a domain to path parameters that the proxy server understands. ✨

So like, container.account.storageurl/object gets translated to container.account.storageurl/path_root/account/container/object, and account.storageurl/path_root/container/object gets translated to account.storageurl/path_root/account/container/object 🦋

  • use — the paste.deploy entry point for the domain_remap middleware. It's a reference to the installed python egg, and bestie, it's normally egg:swift#domain_remap. ✨
  • set log_name — the label used when logging. Default is domain_remap.
  • set log_facility — the syslog log facility. Default is LOG_LOCAL0.
  • set log_level — the logging level. Default is INFO.
  • set log_address — the logging address. Default is /dev/log.
  • set log_headers — turns on the ability to log request headers. Default is False.
  • storage_domain — the domain the middleware is gonna use. You can stack multiple domains here, just separate them with a comma.
  • path_root — the path root value for the storage URL. Default is v1.
  • reseller_prefixes — okay so real talk, browsers can lowercase a host header 💀, so this checks that the reseller prefix on the account is the correct case. It does that by comparing the items in the reseller_prefixes config option to the prefix it actually found. If they match except for case, the item from reseller_prefixes gets used instead of the found reseller prefix. When none of them match, the default reseller prefix is used. And when there's no default reseller prefix configured, any request with an account prefix that's not in that list just gets ignored by this middleware. Defaults to 'AUTH'.
  • default_reseller_prefix — the default reseller prefix, used when none of the configured reseller_prefixes match. When it's not set, no reseller prefix gets added.

🦋 [filter:catch_errors] 🦋

  • use — the paste.deploy entry point for the catch_errors middleware. Reference to the installed python egg, normally egg:swift#catch_errors.
  • set log_name — label used when logging. Default is catch_errors.
  • set log_facility — syslog log facility. Default is LOG_LOCAL0.
  • set log_level — logging level. Default is INFO.
  • set log_address — logging address. Default is /dev/log.
  • set log_headers — enables the ability to log request headers. Default is False.

🌸 [filter:cname_lookup] 🌸

Note: this middleware requires python-dnspython — no cap, you need it.
  • use — the paste.deploy entry point for the cname_lookup middleware. Reference to the installed python egg, normally egg:swift#cname_lookup.
  • set log_name — label used when logging. Default is cname_lookup.
  • set log_facility — syslog log facility. Default is LOG_LOCAL0.
  • set log_level — logging level. Default is INFO.
  • set log_address — logging address. Default is /dev/log.
  • set log_headers — enables the ability to log request headers. Default is False.
  • storage_domain — the domain the middleware is gonna use.
  • lookup_depth — how deep down the CNAME chain it should look for something that matches the storage domain. Default is 1.
  • nameservers — specifies the nameservers used to do the CNAME resolution. If it's unset, the system configuration gets used. You can list multiple nameservers, comma-separated. Default is unset.

🎀 [filter:staticweb] 🎀

Note: put staticweb just after your auth filter(s) in the pipeline. Order matters, fr fr.
  • use — the paste.deploy entry point for the staticweb middleware. Reference to the installed python egg, normally egg:swift#staticweb.
  • set log_name — label used when logging. Default is staticweb.
  • set log_facility — syslog log facility. Default is LOG_LOCAL0.
  • set log_level — logging level. Default is INFO.
  • set log_address — logging address. Default is /dev/log.
  • set log_headers — enables the ability to log request headers. Default is False.

💅 [filter:tempurl] 💅

Note: put tempurl before slo, dlo, and your auth filter(s) in the pipeline. Do not skip this, bestie.
  • use — the paste.deploy entry point for the tempurl middleware. Reference to the installed python egg, normally egg:swift#tempurl.
  • methods — the methods allowed with Temp URLs. Default is 'GET HEAD PUT POST DELETE'.
  • incoming_remove_headers — the headers to remove from incoming requests. It's just a whitespace-delimited list of header names, and names can optionally end with '*' to indicate a prefix match. incoming_allow_headers is the list of exceptions to these removals.
  • incoming_allow_headers — the headers allowed as exceptions to incoming_remove_headers. Again, just a whitespace-delimited list of header names, and names can optionally end with '*' to mean a prefix match.
  • outgoing_remove_headers — the headers to remove from outgoing responses. Whitespace-delimited list of header names, names can optionally end with '*' for a prefix match. outgoing_allow_headers is the list of exceptions to these removals.
  • outgoing_allow_headers — the headers allowed as exceptions to outgoing_remove_headers. Whitespace-delimited list of header names, names can optionally end with '*' for a prefix match.

🍓 [filter:formpost] 🍓

Note: put formpost just before your auth filter(s) in the pipeline.
  • use — the paste.deploy entry point for the formpost middleware. Reference to the installed python egg, normally egg:swift#formpost.

🧚 [filter:name_check] 🧚

Note: it just needs to be placed before the proxy-server in the pipeline.
  • use — the paste.deploy entry point for the name_check middleware. Reference to the installed python egg, normally egg:swift#name_check.
  • forbidden_chars — characters that will not be allowed in a name. The default is:
'"`<>
  • maximum_length — the maximum number of characters that can be in the name. Default is 255.
  • forbidden_regexp — Python regular expressions of substrings that will not be allowed in a name. Default is /./|/../|/.$|/..$/.

⭐ [filter:list-endpoints] ⭐

  • use — the paste.deploy entry point for the list_endpoints middleware. Reference to the installed python egg, normally egg:swift#list_endpoints.
  • list_endpoints_path — default is '/endpoints/'.

👑 [filter:proxy-logging] 👑

Logging for the proxy server now lives in this middleware, okay? And if the access_* variables aren't set, logging directives from [DEFAULT] without "access_" will be used. Slay.

  • use — the paste.deploy entry point for the proxy_logging middleware. Reference to the installed python egg, normally egg:swift#proxy_logging.
  • access_log_name — label used when logging. Default is proxy-server.
  • access_log_facility — syslog log facility. Default is LOG_LOCAL0.
  • access_log_level — logging level. Default is INFO.
  • access_log_address — default is /dev/log.
  • access_log_udp_host — if it's set, access_log_udp_host overrides access_log_address. Default is unset.
  • access_log_udp_port — default is 514.
  • access_log_statsd_host — you can use log_statsd_* from [DEFAULT], or override them right here. It's the StatsD server. IPv4/IPv6 addresses and hostnames are supported. If a hostname resolves to both an IPv4 and an IPv6 address, the IPv4 address gets used.
  • access_log_statsd_port — default is 8125.
  • access_log_statsd_default_sample_rate — default is 1.
  • access_log_statsd_sample_rate_factor — default is 1.
  • access_log_statsd_metric_prefix — default is "" (empty-string).
  • access_log_headers — default is False.
  • access_log_headers_only — if access_log_headers is True and access_log_headers_only is set, only these headers get logged. You can define multiple headers as a comma-separated list, like this:
access_log_headers_only = Host, X-Object-Meta-Mtime
  • reveal_sensitive_prefix — by default, the X-Auth-Token is logged. To obscure the value, set reveal_sensitive_prefix to the number of characters to log. So like, if you set it to 12, only the first 12 characters of the token show up in the log. That way an unauthorized peek at the log file won't let someone actually use the token 🙈. But the first 12-ish characters is unique enough that you can still trace/debug token usage. Set it to 0 to suppress the token completely (it gets replaced by '...' in the log). Default is 16 chars.
Note: reveal_sensitive_prefix will not affect the value logged with access_log_headers=True.
  • log_statsd_valid_http_methods — which HTTP methods are allowed for StatsD logging (comma-sep). Request methods not in this list get "BAD_METHOD" for the <verb> portion of the metric. Default is "GET,HEAD,POST,PUT,DELETE,COPY,OPTIONS".
  • log_anonymization_method — the hashing algorithm for anonymization. It has to be one of the algorithms supported by Python's hashlib. Default is MD5.
  • log_anonymization_salt — salt added as a prefix before hashing the value to anonymize. Default is empty (no salt).
  • log_msg_template — the template used to format access logs. All words surrounded by curly brackets get substituted with the appropriate values. 💖
Quick vibe check bestie 💖 — some keywords are literally timestamps in disguise, and you can glow them up into proper date formats with their matching transformers: 'datetime', 'asctime' or 'iso8601'.

Other timestamp transformers that exist: 's', 'ms', 'us' and 'ns' — that's seconds, milliseconds, microseconds and nanoseconds, slay. ✨

You can also use Python's strftime directives as transformers, and the whole roster is: a, A, b, B, c, d, H, I, j, m, M, p, S, U, w, W, x, X, y, Y, Z. Periodt. 💅

And then some keywords map to user data that you can anonymize with the transformer 'anonymized'. Iconic.

Keywords available are:

  • client_ip (anonymizable)
  • remote_addr (anonymizable)
  • method (request method)
  • path (anonymizable)
  • protocol
  • status_int
  • referer (anonymizable)
  • user_agent (anonymizable)
  • auth_token
  • bytes_recvd (number of bytes received)
  • bytes_sent (number of bytes sent)
  • client_etag (anonymizable)
  • transaction_id
  • headers (anonymizable)
  • request_time (difference between start and end timestamps)
  • source
  • log_info
  • start_time (timestamp at the receiving, timestamp)
  • end_time (timestamp at the end of the treatment, timestamp)
  • ttfb (duration between request and first bytes is sent)
  • policy_index
  • account (account name, anonymizable)
  • container (container name, anonymizable)
  • object (object name, anonymizable)
  • pid (PID of the process emitting the log line)

Example, and no I will not be normal about it:

'{client_ip.anonymized} {remote_addr.anonymized} {start_time.iso8601} {end_time.H}:{end_time.M} {method} acc:{account} cnt:{container} obj:{object.anonymized}'

Default:

'{client_ip} {remote_addr} {end_time.datetime} {method} {path} {protocol} {status_int} {referer} {user_agent} {auth_token} {bytes_recvd} {bytes_sent} {client_etag} {transaction_id} {headers} {request_time} {source} {log_info} {start_time} {end_time} {policy_index}'
Warning: a bad log message template will raise an error in initialization. 😭 So like, don't freestyle it, bestie.

🧺 [filter:bulk] 🧺

Note: put this before both ratelimit and auth in the pipeline. The order matters, iykyk.
  • use — entry point for paste.deploy for the bulk middleware. This is the reference to the installed python egg. Normally it's egg:swift#bulk.
  • max_containers_per_extraction — the default is 10000.
  • max_failed_extractions — the default is 1000.
  • max_deletes_per_request — the default is 10000.
  • max_failed_deletes — the default is 1000.
  • yield_frequency — so here's the tea 🫖: to keep a connection active during a potentially long bulk request, Swift may return whitespace prepended to the actual response body. That whitespace will be yielded no more than every yield_frequency seconds. The default is 10.
  • delete_container_retry_count — Note: this parameter is used during a bulk delete of objects and their container. This would frequently fail because it is very likely that all replicated objects have not been deleted by the time the middleware got a successful response. You can configure the number of retries. And the number of seconds to wait between each retry will be 1.5**retry. The default is 0.

✂️ [filter:slo] ✂️

Note: put this after auth and staticweb in the pipeline. ✨
  • use — entry point for paste.deploy for the slo middleware. This is the reference to the installed python egg. Normally it's egg:swift#slo.
  • max_manifest_segments — the default is 1000.
  • max_manifest_size — the default is 2097152.
  • min_segment_size — the default is 1048576.
  • rate_limit_after_segment — start rate-limiting object segments after the Nth segment of a segmented object. The default is 10 segments.
  • rate_limit_segments_per_sec — once segment rate-limiting kicks in for an object, limit segments served to N per second. The default is 1.
  • max_get_time — time limit on GET requests (seconds). The default is 86400.

🍰 [filter:dlo] 🍰

Note: put this after auth and staticweb in the pipeline. And heads up — if you don't put it in the pipeline, it will be inserted for you. 🪄
  • use — entry point for paste.deploy for the dlo middleware. This is the reference to the installed python egg. Normally it's egg:swift#dlo.
  • rate_limit_after_segment — start rate-limiting object segments after the Nth segment of a segmented object. The default is 10 segments.
  • rate_limit_segments_per_sec — once segment rate-limiting kicks in for an object, limit segments served to N per second. The default is 1.
  • max_get_time — time limit on GET requests (seconds). The default is 86400.

📦 [filter:container-quotas] 📦

Note: put this after auth in the pipeline.
  • use — entry point for paste.deploy for the container_quotas middleware. This is the reference to the installed python egg. Normally it's egg:swift#container_quotas.

🗂️ [filter:account-quotas] 🗂️

Note: put this after auth in the pipeline.
  • use — entry point for paste.deploy for the account_quotas middleware. This is the reference to the installed python egg. Normally it's egg:swift#account_quotas.

🚪 [filter:gatekeeper] 🚪

Note: this middleware requires python-dnspython. Don't forget the dependency bestie, it's giving unfinished homework. 🙈
  • use — entry point for paste.deploy for the gatekeeper middleware. This is the reference to the installed python egg. Normally it's egg:swift#gatekeeper.
  • set log_name — label used when logging. The default is gatekeeper.
  • set log_facility — syslog log facility. The default is LOG_LOCAL0.
  • set log_level — logging level. The default is INFO.
  • set log_address — logging address. The default is /dev/log.
  • set log_headers — enables the ability to log request headers. The default is False.

🔄 [filter:container_sync] 🔄

Note: this middleware requires python-dnspython. Again. 🫶
  • use — entry point for paste.deploy for the container_sync middleware. This is the reference to the installed python egg. Normally it's egg:swift#container_sync.
  • allow_full_urls — set this to false if you want to disallow any full URL values to be set for any new X-Container-Sync-To headers. This will keep any new full urls from coming in, but won't change any existing values already in the cluster. Updating those will have to be done manually, as knowing what the true realm endpoint should be cannot always be guessed. The default is true.
  • current — set this to specify this cluster's //realm/cluster as "current" in /info.

🔬 [filter:xprofile] 🔬

Note: put it at the beginning of the pipeline to profile all middleware. But it is safer to put this after healthcheck. Safety first, gorgeous. ✨
  • use — entry point for paste.deploy for the xprofile middleware. This is the reference to the installed python egg. Normally it's egg:swift#xprofile.
  • profile_module — this option enable you to switch profilers which should inherit from python standard profiler. Currently the supported value can be 'cProfile', 'eventlet.green.profile' etc.
  • log_filename_prefix — this prefix will be used to combine process ID and timestamp to name the profile data file. Make sure the executing user has permission to write into this path (missing path segments will be created, if necessary). If you enable profiling in more than one type of daemon, you must override it with an unique value like, the default is /var/log/swift/profile/account.profile.
  • dump_interval — the profile data will be dumped to local disk based on above naming rule in this interval. The default is 5.0.
  • dump_timestamp — be careful, this option will enable profiler to dump data into the file with time stamp which means there will be lots of files piled up in the directory. The default is false.
  • path — this is the path of the URL to access the mini web UI. The default is __profile__.
  • flush_at_shutdown — clear the data when the wsgi server shutdown. The default is false.
  • unwind — unwind the iterator of applications. Default is false.

📝 [filter:versioned_writes] 📝

Note: put this after slo, dlo in the pipeline. If you don't put it in the pipeline, it will be inserted automatically. 🪄
  • use — entry point for paste.deploy for the versioned_writes middleware. This is the reference to the installed python egg. Normally it's egg:swift#versioned_writes.
  • allow_versioned_writes — enables using versioned writes middleware and exposing configuration settings via HTTP GET /info.
WARNING: setting this option bypasses the "allow_versions" option in the container configuration file, which will be eventually deprecated. See documentation for more details. 👑

✨ APP SECTION ✨

This one's flagged by the section name [app:proxy-server]. Below are all the parameters that are allowed to live in here, bestie. 💖

  • use — the entry point for paste.deploy for the proxy server. It's a reference to the installed python egg. Normally it's egg:swift#proxy. 🥚
  • set log_name — the label used when logging. Default is proxy-server.
  • set log_facility — syslog log facility. Default is LOG_LOCAL0.
  • set log_level — logging level. Default is INFO.
  • set log_address — logging address. Default is /dev/log.
  • log_handoffs — logs when handoff locations get used. Default is True. 👀
  • recheck_account_existence — cache timeout in seconds to send memcached for account existence. Default is 60 seconds.
  • recheck_container_existence — cache timeout in seconds to send memcached for container existence. Default is 60 seconds.
  • object_chunk_size — chunk size to read from object servers. Default is 65536.
  • client_chunk_size — chunk size to read from clients. Default is 65536.
  • node_timeout — request timeout to external services. Default is 10 seconds.
  • recoverable_node_timeout — how long the proxy server will wait for an initial response and to read a chunk of data from the object servers while serving GET / HEAD requests. Timeouts from these requests can be recovered from, so setting this lower than node_timeout gives you quicker error recovery while still allowing a longer timeout for the non-recoverable requests (PUTs). Defaults to node_timeout, and you should override it if node_timeout is set to a high number — otherwise client timeouts can fire before the proxy server even gets a chance to retry. 😭
  • conn_timeout — connection timeout to external services. Default is 0.5 seconds.
  • post_quorum_timeout — how long to wait for requests to finish after a quorum has been established. Default is 0.5 seconds.
  • error_suppression_interval — time in seconds that must elapse since the last error for a node to be considered no longer error limited. Default is 60 seconds.
  • error_suppression_limit — error count to consider a node error limited. Default is 10.
  • allow_account_management — whether account PUTs and DELETEs are even callable at all. Set to 'true' and any authorized user may create and delete accounts; set to 'false' and nobody can, not even authorized users. Default is false.
  • account_autocreate — set to 'true' and authorized accounts that don't exist yet within the Swift cluster get automatically created. Default is set to false.
  • max_containers_per_account — set to a positive value, and trying to create a container when the account already has at least this many containers results in a 403 Forbidden. Note: this is a soft limit, meaning a user might exceed the cap for recheck_account_existence before the 403s kick in. 🙈
  • max_containers_whitelist — a comma separated list of account hashes that ignore the max_containers_per_account cap.
  • deny_host_headers — comma separated list of Host headers to which the proxy will deny requests. Default is empty.
  • sorting_method — storage nodes can be chosen at random (shuffle — the default), by using timing measurements (timing), or by using an explicit match (affinity). Using timing measurements may allow for lower overall latency, while using affinity allows for finer control. In both the timing and affinity cases, equally-sorting nodes are still randomly chosen to spread load. The valid values for sorting_method are "affinity", "shuffle", and "timing". 🎯
  • timing_expiry — if the "timing" sorting_method is used, the timings will only be valid for the number of seconds configured by timing_expiry. Default is 300.
  • concurrent_gets — if "on", then use replica count number of threads concurrently during a GET/HEAD and return with the first successful response. In the EC case, this parameter only affects an EC HEAD, since an EC GET behaves differently. Default is "off".
  • concurrency_timeout — controls how long to wait before firing off the next concurrent_get thread. A value of 0 would be fully concurrent; any other number will stagger the firing of the threads. This number should be between 0 and node_timeout. Default is the value of conn_timeout (0.5).
  • request_node_count — set to the number of nodes to contact for a normal request. You can use '* replicas' at the end to have it use the number given times the number of replicas for the ring being used for the request. Default is '2 * replicas'.
  • read_affinity — specifies which backend servers to prefer on reads. The format is a comma separated list of affinity descriptors of the form <selection>=<priority>. The <selection> may be r<N> for selecting nodes in region N, or r<N>z<M> for selecting nodes in region N, zone M. The <priority> value should be a whole number representing the priority given to the selection; lower numbers are higher priority. Default is empty, meaning no preference. 💅

Example: first read from region 1 zone 1, then region 1 zone 2, then anything in region 2, then everything else:

read_affinity = r1z1=100, r1z2=200, r2=300
  • write_affinity — specifies which backend servers to prefer on writes. The format is a comma separated list of affinity descriptors of the form r<N> for region N, or r<N>z<M> for region N, zone M. If this is set, then when handling an object PUT request, some number (see the write_affinity_node_count setting) of local backend servers will be tried before any nonlocal ones. Default is empty, meaning no preference.

Example: try to write to regions 1 and 2 before writing to any other nodes:

write_affinity = r1, r2
  • write_affinity_node_count — the number of local (as governed by the write_affinity setting) nodes to attempt to contact first on writes, before any non-local ones. The value should be an integer number, or use '* replicas' at the end to have it use the number given times the number of replicas for the ring being used for the request. Default is '2 * replicas'.
  • swift_owner_headers — these are the headers whose values will only be shown to swift_owners. The exact definition of a swift_owner is up to the auth system in use, but usually it indicates administrative responsibilities. The default is 'x-container-read, x-container-write, x-container-sync-key, x-container-sync-to, x-account-meta-temp-url-key, x-account-meta-temp-url-key-2, x-container-meta-temp-url-key, x-container-meta-temp-url-key-2, x-account-access-control'. 👑
  • rate_limit_after_segment — start rate-limiting object segments after the Nth segment of a segmented object. Default is 10 segments.
  • rate_limit_segments_per_sec — once segment rate-limiting kicks in for an object, limit segments served to N per second. Default is 1.
  • nice_priority — modify scheduling priority of server processes. Niceness values range from -20 (most favorable to the process) to 19 (least favorable to the process). The default does not modify priority.
  • ionice_class — modify I/O scheduling class of server processes. I/O niceness class values are IOPRIO_CLASS_RT (realtime), IOPRIO_CLASS_BE (best-effort) and IOPRIO_CLASS_IDLE (idle). The default does not modify class and priority. Works only with ionice_priority.
  • ionice_priority — modify I/O scheduling priority of server processes. I/O niceness priority is a number which goes from 0 to 7. The higher the value, the lower the I/O priority of the process. Works only with ionice_class. Ignored if IOPRIO_CLASS_IDLE is set.

📚 DOCUMENTATION 📚

More in depth documentation about the swift-proxy-server and also OpenStack Swift as a whole can be found at https://docs.openstack.org/swift/latest/admin_guide.html and https://docs.openstack.org/swift/latest/. Go read it, bestie, it's giving ✨ knowledge ✨

👀 SEE ALSO 👀

and that's the whole page, we did it!! 🎀 no cap, you now know proxy-server.conf(5) better than like 90% of the internet.

source: Linux · version/date: 8/26/2011 · page: proxy-server.conf(5)

ily, go slay your config 💖🦋✨

📦 where this page comes from