Actions
Server feature-flagsDetails » History » Revision 19
« Previous |
Revision 19/22
(diff)
| Next »
gstrauss, 2024-02-29 12:54
server.feature-flags option¶
Description¶
server-wide feature control for selected features (since 1.4.56)
server.feature-flags = ( "..." => "..." )
Example:server.feature-flags = ( "server.h2proto" => "enable", "server.h2c" => "enable" )
Options¶
| option | description | notes |
|---|---|---|
"server.h2proto" => "enable" |
HTTP/2 support (default: enable (since 1.4.59)) | (since 1.4.56) |
"server.h2c" => "enable" |
HTTP/2 cleartext support (Upgrade: h2c) (default: enable (since 1.4.59))(requires "server.h2proto" => "enable" to enable) |
(since 1.4.56) |
"server.graceful-shutdown-timeout" => 8 |
max num of secs to wait for existing requests to complete during graceful restart (default: 8 (since 1.4.64)) (slow requests are aborted after timeout to allow graceful restart to proceed) |
(since 1.4.56) |
"server.graceful-restart-bg" => "enable" |
immediate graceful restart while existing requests complete in background process (default: disable) Note: feature works with many lighttpd.conf configurations, but not available in all configurations (e.g. available with independent backends; not available if lighttpd runs backend processes with "bin-path"Use with server.systemd-socket-activation = "enable" (needed before 1.4.68)Fix for #3164 is in lighttpd 1.4.66; using latest lighttpd release is recommended |
(since 1.4.56) |
"server.pcre_jit" => "enable" |
use PCRE JIT with regexes (details: see #2361) (default: enable) | (since 1.4.60) |
"server.clock-jump-restart" => 1800 |
trigger graceful restart if clock jumps more than defined threshold (default: 1800 (30 min)) (if triggered, "server.graceful-shutdown-timeout" forced to 5 sec) |
(since 1.4.60) |
"server.errorlog-high-precision" => "disable" |
high-resolution timestamps in errorlog trace (default: disable) | (since 1.4.60) |
"server.metrics-high-precision" => "disable" |
high-resolution timestamps in metrics collection (default: disable) | (since 1.4.65) |
behavior change; compatibility flags
"server.h2-discard-backend-1xx" => "disable" |
discard 1xx responses from backend (other than 101) when client is HTTP/2 (default: disable) default: relay 1xx from backend when client is HTTP/2, e.g. 103 Early Hints |
(since 1.4.56) |
"server.h1-discard-backend-1xx" => "disable" |
discard 1xx responses from backend (other than 101) when client is HTTP/1.x (default: disable) default: relay 1xx from backend when client is HTTP/1.1, e.g. 103 Early Hints |
(since 1.4.56) |
"server.absolute-dir-redirect" => "disable" |
use fully qualified URI in automatic redirect to directory (default: disable) default: use relative url-path (better abstraction) in automatic redirect to directory |
(since 1.4.56) |
"proxy.force-http10" => "disable" |
force proxy requests to backend to use HTTP/1.0 (default: disable) default: send requests to backend using HTTP/1.1 (alternative: mod_proxy proxy.header += ("force-http10" => "enable") for more selective control) |
(since 1.4.56) |
"ssl.session-cache" => "disable" |
server-side TLS session id cache (default: disable) default: use TLS session tickets (stateless on server-side) |
(since 1.4.56) |
"cgi.tempfile-accum" => "enable" |
accumulate request body in single temporary file if not streaming request body to backend (default: enable) default: temporary file is passed as CGI stdin when request body > 64k (can be disabled if backend immediately reads request body into another temporary file, doubling disk space use on a filesystem with limited size) |
(since 1.4.60) |
"chunkqueue.splice" => "enable" |
splice() data into temporary files (Linux) (default: enable) |
(since 1.4.60) |
"http10.range" => "disable" |
allow Range requests sent with HTTP/1.0 (default: disable) Range request header is part of the HTTP/1.1 specification and might be mishandled by HTTP/1.0 proxies |
(since 1.4.65) |
"auth.delay-invalid-creds" => "enable" |
delay response up to 1 sec for HTTP authentication invalid credentials (default: enable) | (since 1.4.65) |
"auth.http-goaway-invalid-creds => "enable" |
send GOAWAY after HTTP authentication invalid credentials (default: enable) (mitigation should be disabled if different clients are multiplexed on same h2 connection) |
(since 1.4.72) |
"server.http-pathinfo" => "enable" |
allow PATH_INFO in request url-path (default: enable) | (since 1.4.72) |
"gw.upgrade-with-request-body" => "disable" |
allow HTTP/1.1 Upgrade requests with non-zero request body (default: disable) | (since 1.4.74) |
Updated by gstrauss 10 months ago · 19 revisions