r4527 - trunk/varnish-cache/bin/varnishd

phk at projects.linpro.no phk at projects.linpro.no
Mon Feb 1 22:08:37 CET 2010


Author: phk
Date: 2010-02-01 22:08:37 +0100 (Mon, 01 Feb 2010)
New Revision: 4527

Modified:
   trunk/varnish-cache/bin/varnishd/varnishd.1
Log:
Evict the entire section about run-time parameters and replace with
a table produced from the CLI parameter tables.

In the future all updates to these descriptions SHALL happen in the
CLI tables in the C source code.

Fixes	#419



Modified: trunk/varnish-cache/bin/varnishd/varnishd.1
===================================================================
--- trunk/varnish-cache/bin/varnishd/varnishd.1	2010-02-01 21:06:30 UTC (rev 4526)
+++ trunk/varnish-cache/bin/varnishd/varnishd.1	2010-02-01 21:08:37 UTC (rev 4527)
@@ -485,296 +485,846 @@
 use when they arrived.
 .El
 .Ss Run-Time Parameters
-.\" Keep in synch with parspec[] in mgt_param.c and WRK_parspec[] in
- \" mgt_pool.c
 .Pp
-Note that if an run-time parameter is marked
-.Bq experimental ,
-we do not know yet if it is a good idea to change this parameter, or
-if the default value is even sensible.  Caution is advised, and
-feedback is most welcome.
+Runtime parameters are marked with shorthand flags to avoid repeating
+the same text over and over in the table below.
+The meaning of the flags are:
 .Bl -tag -width 4n
+.It Va experimental
+We have no solid information about good/bad/optimal values for this
+parameter.
+Feedback with experience and observations are most welcome.
+.It Va delayed
+This parameter can be changed on the fly, but will not take effect
+immediately.
+.It Va restart
+The worker process must be stopped and restarted, before this parameter
+takes effect.
+.It Va reload
+The VCL programs must be reloaded for this parameter to take effect.
+.El
+.Pp
+Here is a list of all parameters, current as of last time we remembered
+to update the manual page.
+This text is produced from the same text you will find in the CLI
+if you use the
+.Cm param.show
+command, so should there be a new parameter which is not listed here,
+you can find the description using the CLI commands.
+.Pp
+.\" ----------------------------------------------------------------
+.\" This table is generated by running a -DDIAGNOSTIC version of
+.\" varnishd with the secret "-x dumpmdoc" argument
+.\" XXX: Much smarter if we could use .so params.mdoc...
+.\" ----------------------------------------------------------------
+.Bl -tag -width 4n
+.It Va acceptor_sleep_decay
+Default:
+.Dv 0.900
+.br
+Flags:
+.Dv "experimental"
+.br
+.Pp
+If we run out of resources, such as file descriptors or worker threads, the acceptor will sleep between accepts.
+.br
+This parameter (multiplicatively) reduce the sleep duration for each succesfull accept. (ie: 0.9 = reduce by 10%)
+.Pp
+.It Va acceptor_sleep_incr
+Units:
+.Dv s
+.br
+Default:
+.Dv 0.001
+.br
+Flags:
+.Dv "experimental"
+.br
+.Pp
+If we run out of resources, such as file descriptors or worker threads, the acceptor will sleep between accepts.
+.br
+This parameter control how much longer we sleep, each time we fail to accept a new connection.
+.Pp
+.It Va acceptor_sleep_max
+Units:
+.Dv s
+.br
+Default:
+.Dv 0.050
+.br
+Flags:
+.Dv "experimental"
+.br
+.Pp
+If we run out of resources, such as file descriptors or worker threads, the acceptor will sleep between accepts.
+.br
+This parameter limits how long it can sleep between attempts to accept new connections.
+.Pp
 .It Va auto_restart
-Whether to automatically restart the child process if it dies.
+Units:
+.Dv bool
+.br
+Default:
+.Dv on
+.br
 .Pp
-The default is
-.Dv on .
-.It Va backend_http11
-Whether to force the use of HTTP/1.1 when requesting documents from
-the server, or just use the same protocol version as the client which
-triggered the retrieval.
+Restart child process automatically if it dies.
 .Pp
-The default is
-.Dv off .
+.It Va ban_lurker_sleep
+Units:
+.Dv s
+.br
+Default:
+.Dv 0.0
+.br
+.Pp
+How long time does the ban lurker thread sleeps between successfull attempts to push the last item up the purge  list.  It always sleeps a second when nothing can be done.
+.br
+A value of zero disables the ban lurker.
+.Pp
 .It Va between_bytes_timeout
-Default timeout between bytes when receiving data from backend.
-We only wait for this many seconds between bytes before giving up.
-A value of 0 means it will never time out.
-VCL can override this default value for each backend and backend request.
-This parameter does not apply to pipe.
+Units:
+.Dv s
+.br
+Default:
+.Dv 60
+.br
 .Pp
-The default is
-.Dv 60 seconds
-.It Va cli_banner
-Whether to print a banner whenever a client connects to the CLI interface.
+Default timeout between bytes when receiving data from backend. We only wait for this many seconds between bytes before giving up. A value of 0 means it will never time out. VCL can override this default value for each backend request and backend request. This parameter does not apply to pipe.
 .Pp
-The default is
-.Dv on .
+.It Va cache_vbe_conns
+Units:
+.Dv bool
+.br
+Default:
+.Dv off
+.br
+Flags:
+.Dv "experimental"
+.br
+.Pp
+Cache vbe_conn's or rely on malloc, that's the question.
+.Pp
+.It Va cc_command
+Default:
+.Dv exec cc -fpic -shared -Wl,-x -o %o %s
+.br
+Flags:
+.Dv "must_reload"
+.br
+.Pp
+Command used for compiling the C source code to a dlopen(3) loadable object.  Any occurrence of %s in the string will be replaced with the source file name, and %o will be replaced with the output file name.
+.Pp
+.It Va cli_buffer
+Units:
+.Dv bytes
+.br
+Default:
+.Dv 8192
+.br
+.Pp
+Size of buffer for CLI input.
+.br
+You may need to increase this if you have big VCL files and use the vcl.inline CLI command.
+.br
+NB: Must be specified with -p to have effect.
+.Pp
+.It Va cli_timeout
+Units:
+.Dv seconds
+.br
+Default:
+.Dv 10
+.br
+.Pp
+Timeout for the childs replies to CLI requests from the master.
+.Pp
 .It Va client_http11
-Whether to force the use of HTTP/1.1 when responding to client
-requests, or just use the same protocol version as that used by the
-backend server which delivered the requested document.
+Units:
+.Dv bool
+.br
+Default:
+.Dv off
+.br
+Flags:
+.Dv "experimental"
+.br
 .Pp
-The default is
-.Dv off .
+Force all client responses to be HTTP/1.1.
+.br
+By default we copy the protocol version from the backend response.
+.Pp
+.It Va clock_skew
+Units:
+.Dv s
+.br
+Default:
+.Dv 10
+.br
+.Pp
+How much clockskew we are willing to accept between the backend and our own clock.
+.Pp
 .It Va connect_timeout
-Default connection timeout for backend connections.
-We only try to connect to the backend for this many seconds before giving up.
-VCL can override this default value for each backend and backend request.
+Units:
+.Dv s
+.br
+Default:
+.Dv 0.4
+.br
 .Pp
-The default is
-.Dv 0.4 seconds
+Default connection timeout for backend connections. We only try to connect to the backend for this many seconds before giving up. VCL can override this default value for each backend and backend request.
+.Pp
+.It Va default_grace
+Default:
+.Dv 10seconds
+.br
+Flags:
+.Dv "delayed"
+.br
+.Pp
+Default grace period.  We will deliver an object this long after it has expired, provided another thread is attempting to get a new copy.
+.Pp
 .It Va default_ttl
-The default time-to-live assigned to objects if neither the backend
-nor the configuration assign one.
-Note that changes to this parameter are not applied retroactively.
+Units:
+.Dv seconds
+.br
+Default:
+.Dv 120
+.br
 .Pp
-The default is 120 seconds.
+The TTL assigned to objects if neither the backend nor the VCL code assigns one.
+.br
+Objects already cached will not be affected by changes made until they are fetched from the backend again.
+.br
+To force an immediate effect at the expense of a total flush of the cache use "purge.url ."
+.Pp
+.It Va diag_bitmap
+Units:
+.Dv bitmap
+.br
+Default:
+.Dv 0
+.br
+.Pp
+Bitmap controlling diagnostics code:
+.br
+  0x00000001 - CNT_Session states.
+.br
+  0x00000002 - workspace debugging.
+.br
+  0x00000004 - kqueue debugging.
+.br
+  0x00000008 - mutex logging.
+.br
+  0x00000010 - mutex contests.
+.br
+  0x00000020 - waiting list.
+.br
+  0x00000040 - object workspace.
+.br
+  0x00001000 - do not core-dump child process.
+.br
+  0x00002000 - only short panic message.
+.br
+  0x00004000 - panic to stderr.
+.br
+  0x00008000 - panic to abort2().
+.br
+  0x00010000 - synchronize shmlog.
+.br
+  0x00020000 - synchronous start of persistence.
+.br
+  0x80000000 - do edge-detection on digest.
+.br
+Use 0x notation and do the bitor in your head :-)
+.Pp
 .It Va err_ttl
-The default time-to-live assigned to the synthesized error pages.
+Units:
+.Dv seconds
+.br
+Default:
+.Dv 0
+.br
 .Pp
-The default is 0 seconds.
+The TTL assigned to the synthesized error pages
+.Pp
+.It Va esi_syntax
+Units:
+.Dv bitmap
+.br
+Default:
+.Dv 0
+.br
+.Pp
+Bitmap controlling ESI parsing code:
+.br
+  0x00000001 - Don't check if it looks like XML
+.br
+  0x00000002 - Ignore non-esi elements
+.br
+  0x00000004 - Emit parsing debug records
+.br
+Use 0x notation and do the bitor in your head :-)
+.Pp
 .It Va fetch_chunksize
-The default chunk size used when retrieving documents for which the
-backend server does not specify a content length.
+Units:
+.Dv kilobytes
+.br
+Default:
+.Dv 128
+.br
+Flags:
+.Dv "experimental"
+.br
 .Pp
-The default is 128 kilobytes.
+The default chunksize used by fetcher. This should be bigger than the majority of objects with short TTLs.
+.br
+Internal limits in the storage_file module makes increases above 128kb a dubious idea.
+.Pp
 .It Va first_byte_timeout
-Default timeout for receiving first byte from backend.
-We only wait for this many seconds for the first byte before giving up.
-A value of 0 means it will never time out.
-VCL can override this default value for each backend and backend request.
-This parameter does not apply to pipe.
+Units:
+.Dv s
+.br
+Default:
+.Dv 60
+.br
 .Pp
-The default is
-.Dv 60 seconds
+Default timeout for receiving first byte from backend. We only wait for this many seconds for the first byte before giving up. A value of 0 means it will never time out. VCL can override this default value for each backend and backend request. This parameter does not apply to pipe.
+.Pp
 .It Va group
-The name of an unprivileged group to which the child process should
-switch before it starts accepting connections.
-Note that setting
-.Va user
-will automatically set
-.Va group
-to the primary group of the specified user, so if both
-.Va user
-and
-.Va group
-are specified, the latter should be specified last.
+Default:
+.Dv 
+.br
+Flags:
+.Dv "must_restart"
+.br
 .Pp
-The default is "nogroup".
-.It Va sess_workspace
-The size of the per-session workspace for HTTP protocol data.
-For performance reasons, this space is preallocated, so any change to
-this parameter will only apply to new client sessions.
+The unprivileged group to run as.
 .Pp
-The default is 8192 bytes.
+.It Va http_headers
+Units:
+.Dv header lines
+.br
+Default:
+.Dv 64
+.br
+.Pp
+Maximum number of HTTP headers we will deal with.
+.br
+This space is preallocated in sessions and workthreads only objects allocate only space for the headers they store.
+.Pp
 .It Va listen_address
-The address at which to listen for client connections.
-Changes to this parameter will only take effect when the child process
-is restarted.
+Default:
+.Dv :80
+.br
+Flags:
+.Dv "must_restart"
+.br
 .Pp
-The default is 0.0.0.0:80.
+Whitespace separated list of network endpoints where Varnish will accept requests.
+.br
+Possible formats: host, host:port, :port
+.Pp
 .It Va listen_depth
-The depth of the TCP listen queue.
+Units:
+.Dv connections
+.br
+Default:
+.Dv 1024
+.br
+Flags:
+.Dv "must_restart"
+.br
 .Pp
-The default is 512.
+Listen queue depth.
+.Pp
+.It Va log_hashstring
+Units:
+.Dv bool
+.br
+Default:
+.Dv off
+.br
+.Pp
+Log the hash string to shared memory log.
+.Pp
+.It Va log_local_address
+Units:
+.Dv bool
+.br
+Default:
+.Dv off
+.br
+.Pp
+Log the local address on the TCP connection in the SessionOpen shared memory record.
+.Pp
+.It Va lru_interval
+Units:
+.Dv seconds
+.br
+Default:
+.Dv 2
+.br
+Flags:
+.Dv "experimental"
+.br
+.Pp
+Grace period before object moves on LRU list.
+.br
+Objects are only moved to the front of the LRU list if they have not been moved there already inside this timeout period.  This reduces the amount of lock operations necessary for LRU list access.
+.Pp
+.It Va max_esi_includes
+Units:
+.Dv includes
+.br
+Default:
+.Dv 5
+.br
+.Pp
+Maximum depth of esi:include processing.
+.Pp
+.It Va max_restarts
+Units:
+.Dv restarts
+.br
+Default:
+.Dv 4
+.br
+.Pp
+Upper limit on how many times a request can restart.
+.br
+Be aware that restarts are likely to cause a hit against the backend, so don't increase thoughtlessly.
+.Pp
+.It Va obj_workspace
+Units:
+.Dv bytes
+.br
+Default:
+.Dv 0
+.br
+Flags:
+.Dv "delayed"
+.br
+.Pp
+Bytes of HTTP protocol workspace allocated for objects. This space must be big enough for the entire HTTP protocol header and any edits done to it in the VCL code while it is cached.
+.br
+Minimum is 1024 bytes.
+.Pp
 .It Va overflow_max
-The maximum depth of the overflow queue as a percentage of
-.Va thread_pool_max .
-The overflow queue holds requests waiting for a worker thread to
-become available.
-If the overflow queue fills up,
-.Nm
-will start dropping new connections.
+Units:
+.Dv %
+.br
+Default:
+.Dv 100
+.br
+Flags:
+.Dv "experimental"
+.br
 .Pp
-The default is 100%.
+Percentage permitted overflow queue length.
+.Pp
+This sets the ratio of queued requests to worker threads, above which sessions will be dropped instead of queued.
+.Pp
 .It Va ping_interval
-The interval at which the parent process will ping the child process
-to ascertain that it is still present and functioning.
+Units:
+.Dv seconds
+.br
+Default:
+.Dv 3
+.br
+Flags:
+.Dv "must_restart"
+.br
 .Pp
-The default is 3 seconds.
+Interval between pings from parent to child.
+.br
+Zero will disable pinging entirely, which makes it possible to attach a debugger to the child.
+.Pp
 .It Va pipe_timeout
-The time to wait before dropping an idle pipe mode connection.
+Units:
+.Dv seconds
+.br
+Default:
+.Dv 60
+.br
 .Pp
-The default is 60 seconds.
+Idle timeout for PIPE sessions. If nothing have been received in either direction for this many seconds, the session is closed.
+.Pp
+.It Va prefer_ipv6
+Units:
+.Dv bool
+.br
+Default:
+.Dv off
+.br
+.Pp
+Prefer IPv6 address when connecting to backends which have both IPv4 and IPv6 addresses.
+.Pp
+.It Va purge_dups
+Units:
+.Dv bool
+.br
+Default:
+.Dv on
+.br
+.Pp
+Detect and eliminate duplicate purges.
+.Pp
 .It Va rush_exponent
-.Bq experimental
+Units:
+.Dv requests per request
+.br
+Default:
+.Dv 3
+.br
+Flags:
+.Dv "experimental"
+.br
 .Pp
 How many parked request we start for each completed request on the object.
+.br
+NB: Even with the implict delay of delivery, this parameter controls an exponential increase in number of worker threads.  
 .Pp
-NB: Even with the implict delay of delivery, this parameter controls
-an exponential increase in number of worker threads.
-.Pp
-The default is 3 requests per request.
 .It Va saintmode_threshold
-The number of items allowed on the saintmode list for a backend before it
-is considered sick. A value of 0 disables saintmode and any associated
-overhead.
+Units:
+.Dv objects
+.br
+Default:
+.Dv 10
+.br
+Flags:
+.Dv "experimental"
+.br
 .Pp
-The default is 10 objects
+The maximum number of objects held off by saint mode before no further will be made to the backend until one times out.  A value of 0 disables saintmode.
+.Pp
+.It Va send_timeout
+Units:
+.Dv seconds
+.br
+Default:
+.Dv 600
+.br
+Flags:
+.Dv "delayed"
+.br
+.Pp
+Send timeout for client connections. If no data has been sent to the client in this many seconds, the session is closed.
+.br
+See setsockopt(2) under SO_SNDTIMEO for more information.
+.Pp
 .It Va sendfile_threshold
-The size threshold beyond which documents are sent to the client using
-.Xr sendfile 2
-instead of
-.Xr writev 2 .
-This is not likely to have any effect unless the working set is too
-large to fit in physical memory.
+Units:
+.Dv bytes
+.br
+Default:
+.Dv -1
+.br
+Flags:
+.Dv "experimental"
+.br
 .Pp
-Note that several operating systems have known bugs which make it
-inadvisable to use this.
+The minimum size of objects transmitted with sendfile.
 .Pp
-The default is \-1, which disables the use of sendfile altogether.
-.It Va send_timeout
-The time to wait before dropping the connection to a client which is
-not accepting data sent to it.
-.Pp
-The default is 600 seconds.
 .It Va sess_timeout
-The time to wait before dropping an idle client session.
+Units:
+.Dv seconds
+.br
+Default:
+.Dv 5
+.br
 .Pp
-The default is 5 seconds.
-.It Va srcaddr_hash
-The size of the hash table used to store per-client accounting data.
-This should preferably be a prime number.
+Idle timeout for persistent sessions. If a HTTP request has not been received in this many seconds, the session is closed.
 .Pp
-The default is 1049.
-.It Va srcaddr_ttl
-The length of time to keep per-client accounting records.
-Setting this to 0 will disable per-client accounting.
+.It Va sess_workspace
+Units:
+.Dv bytes
+.br
+Default:
+.Dv 16384
+.br
+Flags:
+.Dv "delayed"
+.br
+.Pp
+Bytes of HTTP protocol workspace allocated for sessions. This space must be big enough for the entire HTTP protocol header and any edits done to it in the VCL code.
+.br
+Minimum is 1024 bytes.
+.Pp
+.It Va session_linger
+Units:
+.Dv ms
+.br
+Default:
+.Dv 50
+.br
+Flags:
+.Dv "experimental"
+.br
+.Pp
+How long time the workerthread lingers on the session to see if a new request appears right away.
+.br
+If sessions are reused, as much as half of all reuses happen within the first 100 msec of the previous request completing.
+.br
+Setting this too high results in worker threads not doing anything for their keep, setting it too low just means that more sessions take a detour around the waiter.
+.Pp
+.It Va session_max
+Units:
+.Dv sessions
+.br
+Default:
+.Dv 100000
+.br
+.Pp
+Maximum number of sessions we will allocate before just dropping connections.
+.br
+This is mostly an anti-DoS measure, and setting it plenty high should not hurt, as long as you have the memory for it.
+.Pp
+.It Va shm_reclen
+Units:
+.Dv bytes
+.br
+Default:
+.Dv 255
+.br
+.Pp
+Maximum number of bytes in SHM log record.
+.br
+Maximum is 65535 bytes.
+.Pp
+.It Va shm_workspace
+Units:
+.Dv bytes
+.br
+Default:
+.Dv 8192
+.br
+Flags:
+.Dv "delayed"
+.br
+.Pp
+Bytes of shmlog workspace allocated for worker threads. If too big, it wastes some ram, if too small it causes needless flushes of the SHM workspace.
+.br
+These flushes show up in stats as "SHM flushes due to overflow".
+.br
+Minimum is 4096 bytes.
+.Pp
+.It Va syslog_cli_traffic
+Units:
+.Dv bool
+.br
+Default:
+.Dv on
+.br
+.Pp
+Log all CLI traffic to syslog(LOG_INFO).
+.Pp
 .It Va thread_pool_add_delay
-.Bq experimental
+Units:
+.Dv milliseconds
+.br
+Default:
+.Dv 20
+.br
+Flags:
+.Dv "experimental"
+.br
 .Pp
-Wait at least many miliseconds between creating threads.
+Wait at least this long between creating threads.
 .Pp
 Setting this too long results in insuffient worker threads.
 .Pp
-Settings this too high increase the risk for worker thread pile-up.
+Setting this too short increases the risk of worker thread pile-up.
 .Pp
-The default is 20 miliseconds.
 .It Va thread_pool_add_threshold
-.Bq experimental
+Units:
+.Dv requests
+.br
+Default:
+.Dv 2
+.br
+Flags:
+.Dv "experimental"
+.br
 .Pp
-Overflow threshold in number of requests for worker thread creation.
+Overflow threshold for worker thread creation.
 .Pp
-Setting this too low, will result in excess worker threads, which is
-generally a bad idea.
+Setting this too low, will result in excess worker threads, which is generally a bad idea.
 .Pp
 Setting it too high results in insuffient worker threads.
 .Pp
-The default is 2 requests
 .It Va thread_pool_fail_delay
-.Bq experimental
+Units:
+.Dv milliseconds
+.br
+Default:
+.Dv 200
+.br
+Flags:
+.Dv "experimental"
+.br
 .Pp
-Wait at least this many miliseconds after a failed thread creation
-before trying to create another thread.
+Wait at least this long after a failed thread creation before trying to create another thread.
 .Pp
-Failure to create a worker thread is often a sign that the end is
-near, because the process is running out of RAM resources for thread
-stacks.  This delay tries to not rush it on needlessly.
+Failure to create a worker thread is often a sign that  the end is near, because the process is running out of RAM resources for thread stacks.
+.br
+This delay tries to not rush it on needlessly.
 .Pp
-If thread creation failures are a problem, check that
-.Va thread_pool_max
-is not too high.
+If thread creation failures are a problem, check that thread_pool_max is not too high.
 .Pp
-It may also help to increase
-.Va thread_pool_timeout
-and
-.Va thread_pool_min ,
-to reduce the rate at which treads are destroyed and later recreated.
+It may also help to increase thread_pool_timeout and thread_pool_min, to reduce the rate at which treads are destroyed and later recreated.
 .Pp
-The default is 200 miliseconds
 .It Va thread_pool_max
-.Bq experimental
+Units:
+.Dv threads
+.br
+Default:
+.Dv 500
+.br
+Flags:
+.Dv "delayed, experimental"
+.br
 .Pp
-The maximum total number of worker threads.
-If the number of concurrent requests rises beyond this number,
-requests will start queueing up waiting for a worker thread to pick
-them up.
-Higher values may improve response time but will increase pressure on
-the scheduler.
+The maximum number of worker threads in all pools combined.
 .Pp
-The default is 1000.
+Do not set this higher than you have to, since excess worker threads soak up RAM and CPU and generally just get in the way of getting work done.
+.Pp
 .It Va thread_pool_min
-.Bq experimental
+Units:
+.Dv threads
+.br
+Default:
+.Dv 5
+.br
+Flags:
+.Dv "delayed, experimental"
+.br
 .Pp
-The minimum total number of worker threads.
-Higher values may allow
-.Nm
-to respond faster to a sudden increase in traffic.
+The minimum number of threads in each worker pool.
 .Pp
-The default is 5.
+Increasing this may help ramp up faster from low load situations where threads have expired.
+.Pp
+Minimum is 2 threads.
+.Pp
 .It Va thread_pool_purge_delay
-.Bq experimental
+Units:
+.Dv milliseconds
+.br
+Default:
+.Dv 1000
+.br
+Flags:
+.Dv "delayed, experimental"
+.br
 .Pp
-Wait this many miliseconds between purging threads.
+Wait this long between purging threads.
 .Pp
-This controls the decay of thread pool when idle(-ish).
+This controls the decay of thread pools when idle(-ish).
 .Pp
-The minimum is 100 miliseconds, the default is 1000 miliseconds.
+Minimum is 100 milliseconds.
+.Pp
 .It Va thread_pool_stack
-.Bq experimental
+Units:
+.Dv bytes
+.br
+Default:
+.Dv -1
+.br
+Flags:
+.Dv "experimental"
+.br
 .Pp
-Worker thread stack size.  The number used is the amount of bytes used
-as thread stack size.
+Worker thread stack size.  In particular on 32bit systems you may need to tweak this down to fit many threads into the limited address space.
 .Pp
-On 32bit systems in particular you may need to tweak this down to fit
-many threads into the limited address space.
 .It Va thread_pool_timeout
-.Bq experimental
+Units:
+.Dv seconds
+.br
+Default:
+.Dv 300
+.br
+Flags:
+.Dv "delayed, experimental"
+.br
 .Pp
-The amount of time a worker thread can be idle before it is killed,
-when the number of worker threads exceeds
-.Va thread_pool_min .
+Thread idle threshold.
 .Pp
-The default is 120 seconds.
+Threads in excess of thread_pool_min, which have been idle for at least this long are candidates for purging.
+.Pp
+Minimum is 1 second.
+.Pp
 .It Va thread_pools
-The number of worker thread pools.
-Higher values reduce lock contention but increase pressure on the
-scheduler.
-Note that a decrease of this parameter will only take effect after a
-restart.
+Units:
+.Dv pools
+.br
+Default:
+.Dv 2
+.br
+Flags:
+.Dv "delayed, experimental"
+.br
 .Pp
-The default is 2.
+Number of worker thread pools.
+.Pp
+Increasing number of worker pools decreases lock contention.
+.Pp
+Too many pools waste CPU and RAM resources, and more than one pool for each CPU is probably detrimal to performance.
+.Pp
+Can be increased on the fly, but decreases require a restart to take effect.
+.Pp
 .It Va thread_stats_rate
-.Bq experimental
+Units:
+.Dv requests
+.br
+Default:
+.Dv 10
+.br
+Flags:
+.Dv "experimental"
+.br
 .Pp
-Worker threads accumulate statistics, and dump these into the global
-stats counters if the lock is free when they finish a request.
+Worker threads accumulate statistics, and dump these into the global stats counters if the lock is free when they finish a request.
+.br
+This parameters defines the maximum number of requests a worker thread may handle, before it is forced to dump its accumulated stats into the global counters.
 .Pp
-This parameters defines the maximum number of requests a worker thread
-may handle, before it is forced to dump its accumulated stats into the
-global counters.
 .It Va user
-The name of an unprivileged user to which the child process should
-switch before it starts accepting connections.
-Note that setting
-.Va user
-will automatically set
-.Va group
-to the primary group of the specified user, so if both
-.Va user
-and
-.Va group
-are specified, the latter should be specified last.
+Default:
+.Dv 
+.br
+Flags:
+.Dv "must_restart"
+.br
 .Pp
-The default is "nobody".
+The unprivileged user to run as.  Setting this will also set "group" to the specified user's primary group.
+.Pp
 .It Va vcl_trace
-Whether to issue log entries for calls to VCL code and their results.
-Note that this will generate large amounts of log data.
+Units:
+.Dv bool
+.br
+Default:
+.Dv off
+.br
 .Pp
-The default is
-.Dv off .
+Trace VCL execution in the shmlog.
+.br
+Enabling this will allow you to see the path each request has taken through the VCL program.
+.br
+This generates a lot of logrecords so it is off by default.
+.Pp
+.It Va waiter
+Default:
+.Dv default
+.br
+Flags:
+.Dv "must_restart, experimental"
+.br
+.Pp
+Select the waiter kernel interface.
+.Pp
 .El
+.\" ----------------------------------------------------------------
+.\" End of machine generated table
+.\" ----------------------------------------------------------------
 .Ss Purge expressions
 A purge expression consists of one or more conditions.
 A condition consists of a field, an operator, and an argument.



More information about the varnish-commit mailing list