Performance Logging iRule (Rule_http_log)
Problem this snippet solves: Here's a logging iRule. You'll need a HSL syslog pool to log too. Various bits gathered from other posts on DevCentral. Sharing in case there is interest. Make sure your rsyslogd is setup to use the newer syslog format like RFC-5424 including milliseconds and timezone info.Includes Country (co) and logs individual request times for each request on a HTTP/1.1 connection. To configure F5 logging to use milliseconds and timezone, disable logging in the gui and use tmsh edit sys syslog and something like: include " # short hostnames options { use_fqdn(no); }; # Remote syslog in RFC5424 - Tim Riker <Tim@Rikers.org> destination remotesyslog { syslog(\"10.1.2.3\" transport(\"udp\") port(51443) ts_format(iso)); }; log { source(s_syslog_pipe); destination(remotesyslog); }; " Uses upvar and proc. Tested on 11.6 - 15.1 This tracks connection info in a table and then copies that down to the per-request log() to handle reporting on http2. This version works around a BIG-IP bug where HTTP::version does not report 2 or higher for http2 and later requests. With http2 profiles, subsequent requests using the same connection can generate this error in the logs if HTTP::respond HTTP::redirect or HTTP::retry is called from and earlier iRule. Reorder your iRules to avoid this. <HTTP_REQUEST> - No HTTP header is cached - ERR_NOT_SUPPORTED (line 1)invoked from within "HTTP::method" How to use this snippet: Add this iRule to whatever virtual hosts you desire. I always add it as the first rule. If you have a rule that sets headers you want to track, you may want this after the rule that sets headers. Interesting Splunk queries can be created like: index=* perflog | timechart avg(cpu_5sec) by host limit=10 to show load across multiple F5s. index=* perflog | timechart max(upstream_time) by http_host limit=10 to show long request times by http_host Any other iRule may add things to the log() array and those will get added to the single hsl output. If you create a dg_http_log datagroup, that will be used to filter what gets logged. Tested on version: 13.0 - 15.1 # Rule_http_log # http logging - Tim Riker <Tim@Rikers.org> # bits taken from this post: # https://devcentral.f5.com/questions/irule-for-getting-total-response-time-server-response-time-and-server-connection-time # iRule performance tracking # https://devcentral.f5.com/questions/Timing-iRules timing on # timing is on by default in 11.5.0+ to see stats: # tmsh show ltm rule Rule_http_log # # if the dg_http_log datagroup exists then vips or hosts/paths in dg_http_log that start with # "NONE" no logging (really anything other than empty) # "INFO" normal logging # "FINE" full request and response headers and CLIENT_CLOSED # # upstream_time := 15000 in the datagroup to log all requests over 15 seconds # # example: # "/Common/vs_www.example.com_HTTPS" := "FINE" - logged including CLIENT_CLOSED # "www.example.com/" := "INFO" - logged # "www.example.com/somepath" := "FINE" - full headers # "www.example.com/otherpath" := "NONE" - not logged when RULE_INIT { # hostname up to first dot set static::hostname [getfield [info hostname] "." 1] } # not calling /Common/proc:hsllog as this logs when the request occurred # instead of the time it calls hsllog at the end of the request proc hsllog {time mylog} { upvar 1 $mylog log # https://tools.ietf.org/html/rfc5424 <local0.info>version rfc-3339time host procid msgid structured_data log # should be able to use a "Z" here instead of "+00:00" but our splunk logs don't handle that # 134 = local0.info set output "<134>1 [clock format [string range $time 0 end-3] -gmt 1 -format %Y-%m-%dT%H:%M:%S.[string range $time end-2 end]+00:00] ${static::hostname} httplog [TMM::cmp_group].[TMM::cmp_unit] - -" foreach key [lsort [array names log]] { if { ($log($key) matches_regex {[\" ;,:]}) } { append output " $key=\"[string map {\" "|"} $log($key)]\"" } else { append output " $key=$log($key)" } } # avoid marking virtual server up when hsl pool is up # https://support.f5.com/csp/article/K14505 set hsl pool_syslog HSL::send [HSL::open -proto UDP -pool $hsl] $output } when CLIENT_ACCEPTED { # calculate and track milliseconds # is this / 1000 guaranteed to be clock seconds? TCL docs say no, but it looks like on f5 it is. set tcp_start_time [clock clicks -milliseconds] set log(loglevel) 0 if { [class exists dg_http_log] } { # virtual name entries need to be full path, ie: /Common/vs_www.example.com_HTTP switch -- [string range [class match -value -- [virtual name] equals dg_http_log] 0 3] { "FINE" { set log(loglevel) 2 } "INFO" { set log(loglevel) 1 } default { set log(loglevel) 0 } } } table set -subtable [IP::client_addr]:[TCP::client_port] loglevel $log(loglevel) table set -subtable [IP::client_addr]:[TCP::client_port] tmm "[TMM::cmp_group].[TMM::cmp_unit]" table set -subtable [IP::client_addr]:[TCP::client_port] client_addr [IP::client_addr] table set -subtable [IP::client_addr]:[TCP::client_port] client_port [TCP::client_port] table set -subtable [IP::client_addr]:[TCP::client_port] cpu_5sec [cpu usage 5secs] table set -subtable [IP::client_addr]:[TCP::client_port] virtual_name [virtual name] set co [whereis [IP::client_addr] country] if { $co eq "" } { set co unknown } table set -subtable [IP::client_addr]:[TCP::client_port] co $co } when HTTP_REQUEST { set http_request_time [clock clicks -milliseconds] set keys [table keys -subtable [IP::client_addr]:[TCP::client_port]] foreach key $keys { set log($key) "[table lookup -subtable "[IP::client_addr]:[TCP::client_port]" "$key"]" } if {[HTTP::has_responded]} { # The rule should come BEFORE any rules that do things like redirects set log(http_has_responded) [HTTP::has_responded] set log(loglevel) 1 set log(event) HTTP_REQUEST call hsllog $http_request_time log return } if { [class exists dg_http_log] } { set logsetting [class match -value -- [HTTP::host][HTTP::uri] starts_with dg_http_log] if { $logsetting ne "" } { # override log(loglevel) if we found something switch -- [string range $logsetting 0 3] { "FINE" { set log(loglevel) 2 } "INFO" { set log(loglevel) 1 } default { set log(loglevel) 0 } } } } set log(http_host) [HTTP::host] set log(http_uri) [HTTP::uri] set log(http_method) [HTTP::method] # request_num might not be accurate for HTTP2 set log(request_num) [HTTP::request_num] set log(request_size) [string length [HTTP::request]] # BUG http2 reported as http1 in pre 16.x # https://cdn.f5.com/product/bugtracker/ID842053.html set log(http_version) [HTTP::version] if { [catch \[HTTP2::version\] result] == 1 } { if { $result contains "Operation not supported" } { #log local0. "HTTP version is: [HTTP::version]" } else { set h2ver [eval "\HTTP2::version"] # we might have http2 support, but not be http2 if { $h2ver != 0 } { set log(http_version) $h2ver } } } #log local0. "http_version = $log(http_version)" if { $log(loglevel) > 1 } { foreach {header} [HTTP::header names] { set log(req-$header) [HTTP::header $header] } } else { foreach {header} {"connection" "content-length" "keep-alive" "last-modified" "policy-cn" "referer" "transfer-encoding" "user-agent" "x-forwarded-for" "x-forwarded-proto" "x-forwarded-scheme"} { if { [HTTP::header exists $header] } { set log(req-$header) [HTTP::header $header] } } } } when LB_SELECTED { set lb_selected_time [clock clicks -milliseconds] set log(server_addr) [LB::server addr] set log(server_port) [LB::server port] set log(pool) [LB::server pool] } when SERVER_CONNECTED { set log(connection_time) [expr {[clock clicks -milliseconds] - $lb_selected_time}] set log(snat_addr) [IP::local_addr] set log(snat_port) [TCP::local_port] } when LB_FAILED { set log(event_info) [event info] } when HTTP_REJECT { set log(http_reject) [HTTP::reject_reason] } when HTTP_REQUEST_SEND { set http_request_send_time [clock clicks -milliseconds] } when HTTP_RESPONSE { set log(upstream_time) [expr {[clock clicks -milliseconds] - $http_request_send_time}] set log(http_status) [HTTP::status] if { $log(loglevel) > 1 } { foreach {header} [HTTP::header names] { set log(res-$header) [HTTP::header $header] } } else { foreach {header} {"cache-control" "connection" "content-encoding" "content-length" "content-type" "content-security-policy" "keep-alive" "last-modified" "location" "server" "www-authenticate"} { if { [HTTP::header exists $header] } { set log(res-$header) [HTTP::header $header] } } } # if logging is off, but upstream_time is over threshold in datagroup, log anyway if { ($log(loglevel) < 1) && [class exists dg_http_log] } { set log_upstream_time [class match -value -- upstream_time equals dg_http_log] if {$log_upstream_time ne "" && $log(upstream_time) >= $log_upstream_time} { set log(over_upstream_time) $log_upstream_time set log(loglevel) 1 } } } when HTTP_RESPONSE_RELEASE { if { [info exists http_request_time] } { set log(http_time) "[expr {[clock clicks -milliseconds] - $http_request_time}]" # push http_time into table so CLIENT_CLOSED can see it in HTTP/2 table set -subtable [IP::client_addr]:[TCP::client_port] http_time $log(http_time) } else { set http_request_time [clock clicks -milliseconds] } set log(event) HTTP_RESPONSE_RELEASE if { $log(loglevel) > 0 } { call hsllog $http_request_time log } } when HTTP_DISABLED { set log(http_passthrough_reason) [HTTP::passthrough_reason] } when CLIENT_CLOSED { # grab log() values from table set keys [table keys -subtable [IP::client_addr]:[TCP::client_port]] foreach key $keys { set log($key) "[table lookup -subtable "[IP::client_addr]:[TCP::client_port]" "$key"]" } set log(tcp_time) "[expr {[clock clicks -milliseconds] - $tcp_start_time}]" set log(event) CLIENT_CLOSED # http_time didn't get set, log here (HTTP_RESPONSE_RELEASE never called, catch redirects, aborted connections) if { not ([info exists log(http_time)]) } { if { [info exists http_request_time] } { # called HTTP_REQUEST but not HTTP_RESPONSE_RELEASE using HTTP 1.0 or 1.1 set log(http_time) "[expr {[clock clicks -milliseconds] - $http_request_time}]" } call hsllog $tcp_start_time log } elseif { $log(loglevel) > 1 } { call hsllog $tcp_start_time log } # clean out table when client disconnects table delete -subtable [IP::client_addr]:[TCP::client_port] -all }
Log Tcp And Http Request Response Info
Problem this snippet solves: This iRule logs a line for the following events: when a new TCP connection is established with a client when the HTTP headers of an HTTP request are received from the client when the HTTP headers of an HTTP response are received from the pool member when the TCP connection with a client is closed Code : # Here is a sample of the log output for a single TCP connection with three HTTP requests: : New TCP connection from 192.168.99.210:2675 to 192.168.101.41:80 : Client 192.168.99.210:2675 -> test_http_vip/test0.html?parameter=val (request) : Client 192.168.99.210:2675 -> test_http_vip/test0.html?parameter=val (response) - pool info http_pool 192.168.101.45 80 - status: 200 (request/response delta: 0ms) : Client 192.168.99.210:2675 -> test_http_vip/test1.html?parameter=val (request) : Client 192.168.99.210:2675 -> test_http_vip/test1.html?parameter=val (response) - pool info http_pool 192.168.101.45 80 - status: 200 (request/response delta: 0ms) : Client 192.168.99.210:2675 -> test_http_vip/test2.html?parameter=val (request) : Client 192.168.99.210:2675 -> test_http_vip/test2.html?parameter=val (response) - pool info http_pool 192.168.101.45 80 - status: 200 (request/response delta: 1ms) : Closed TCP connection from 192.168.99.210:2675 to 192.168.101.41:80 (open for: 1078ms) when CLIENT_ACCEPTED { # Get time for start of TCP connection in milleseconds set tcp_start_time [clock clicks -milliseconds] # Log the start of a new TCP connection log local0. "New TCP connection from [IP::client_addr]:[TCP::client_port] to [IP::local_addr]:[TCP::local_port]" } when HTTP_REQUEST { # Get time for start of HTTP request set http_request_time [clock clicks -milliseconds] # Log the start of a new HTTP request set LogString "Client [IP::client_addr]:[TCP::client_port] -> [HTTP::host][HTTP::uri]" log local0. "$LogString (request)" } when LB_SELECTED { log local0. "Client [IP::client_addr]:[TCP::client_port]: Selected [LB::server]" } when LB_FAILED { log local0. "Client [IP::client_addr]:[TCP::client_port]: Failed to [LB::server]" } when SERVER_CONNECTED { log local0. "Client [IP::client_addr]:[TCP::client_port]: Connected to [IP::server_addr]:[TCP::server_port]" } when HTTP_RESPONSE { # Received the response headers from the server. Log the pool name, IP and port, status and time delta log local0. "$LogString (response) - pool info: [LB::server] - status: [HTTP::status] (request/response delta: [expr {[clock clicks -milliseconds] - $http_request_time}] ms)" } when CLIENT_CLOSED { # Log the end time of the TCP connection log local0. "Closed TCP connection from [IP::client_addr]:[TCP::client_port] to [IP::local_addr]:[TCP::local_port] (open for: [expr {[clock clicks -milliseconds] - $tcp_start_time}] ms)" }Log large HTTP payloads in chunks locally and remotely
Problem this snippet solves: Log HTTP POST request payloads remotely via High Speed Logging (HSL) to a syslog server and locally. Code : # Log POST request payloads remotely via HSL to a syslog server and locally. # Based on Steve Hillier's example and the HTTP::collect wiki page # https://devcentral.f5.com/s/wiki/iRules.http__collect.ashx # Note that although any size payload can theoretically be collected, the maximum size of a Tcl variable in v9 and v10 is 4MB # with a smaller functional maximum after charset expansion of approximately 1Mb. # In v11, the maximum variable size was increased to 32Mb. when RULE_INIT { # Log debug to /var/log/ltm? 1=yes, 0=no set static::payload_dbg 1 # Limit payload collection to 5Mb set static::max_collect_len 5242880 # HSL pool name set static::hsl_pool "my_hsl_tcp_pool" # Max characters to log locally (must be less than 1024 bytes) # https://devcentral.f5.com/s/wiki/iRules.log.ashx set static::max_chars 900 } when HTTP_REQUEST { # Only collect POST request payloads if {[HTTP::method] equals "POST"}{ if {$static::payload_dbg}{log local0. "POST request"} # Open HSL connection set hsl [HSL::open -proto TCP -pool $static::hsl_pool] # Get the content length so we can request the data to be processed in the HTTP_REQUEST_DATA event. if {[HTTP::header exists "Content-Length"]}{ set content_length [HTTP::header "Content-Length"] } else { set content_length 0 } # content_length of 0 indicates chunked data (of unknown size) if {$content_length > 0 && $content_length < $static::max_collect_len}{ set collect_length $content_length } else { set collect_length $static::max_collect_len } if {$static::payload_dbg}{log local0. "Content-Length: $content_length, Collect length: $collect_length"} } } when HTTP_REQUEST_DATA { # Log the bytes collected if {$static::payload_dbg}{log local0. "Collected [HTTP::payload length] bytes"} # Send all the collected payload to the remote syslog server HSL::send $hsl "<190>[HTTP::payload]\n" # Log the payload locally if {[HTTP::payload length] < $static::max_chars}{ log local0. "Payload=[HTTP::payload]" } else { # Initialize variables set remaining $payload set position 0 set count 1 set bytes_logged 0 # Loop through and log each chunk of the payload while {[string length $remaining] > $static::max_chars}{ # Get the current chunk to log (subtract 1 from the end as string range is 0 indexed) set current [string range $remaining $position [expr {$position + $static::max_chars -1}]] log local0. "chunk $count=$current" # Add the length of the current chunk to the position for the next chunk incr position [string length $current] # Get the next chunk to log set remaining [string range $remaining $position end] incr count incr bytes_logged $position log local0. "remaining bytes=[string length $remaining], \$position=$position, \$count=$count, \$bytes_logged=$bytes_logged" } if {[string length $remaining]}{ log local0. "chunk $count=$current" incr bytes_logged [string length $remaining] } log local0. "Logged $count chunks for a total of $bytes_logged bytes" } }
