Author: | Dag-Erling Smørgrav |
---|---|
Author: | Stig Sandbeck Mathisen |
Author: | Per Buer |
Date: | 2010-05-31 |
Version: | 1.0 |
Manual section: | 1 |
The varnishd daemon accepts HTTP requests from clients, passes them on to a backend server and caches the returned documents to better satisfy future requests for the same document.
-C | Print VCL code compiled to C language and exit. Specify the VCL file to compile with the -f option. |
-d | Enables debugging mode: The parent process runs in the foreground with a CLI connection on stdin/stdout, and the child process must be started explicitly with a CLI command. Terminating the parent process will also terminate the child. |
-F | Run in the foreground. |
-f config | Use the specified VCL configuration file instead of the builtin default. See vcl(7) for details on VCL syntax. |
-g group | Specifies the name of an unprivileged group to which the child process should switch before it starts accepting connections. This is a shortcut for specifying the group run-time parameter. |
-i identity | Specify the identity of the varnish server. This can be accessed using server.identity from VCL |
-l shmlogsize | Specify size of shmlog file. Scaling suffixes like ‘k’, ‘m’ can be used up to (e)tabytes. Default is 80 Megabytes. Specifying less than 8 Megabytes is unwise. |
-n name | Specify a name for this instance. Amonst other things, this name is used to construct the name of the directory in which varnishd keeps temporary files and persistent state. If the specified name begins with a forward slash, it is interpreted as the absolute path to the directory which should be used for this purpose. |
-P file | Write the process’s PID to the specified file. |
-S file | Path to a file containing a secret used for authorizing access to the management port. |
-t ttl | Specifies a hard minimum time to live for cached documents. This is a shortcut for specifying the default_ttl run-time parameter. |
-u user | Specifies the name of an unprivileged user to which the child process should switch before it starts accepting connections. This is a shortcut for specifying the user run- time parameter. If specifying both a user and a group, the user should be specified first. |
-V | Display the version number and exit. |
-w min[,max[,timeout]]
Start at least min but no more than max worker threads with the specified idle timeout. This is a shortcut for specifying the thread_pool_min, thread_pool_max and thread_pool_timeout run-time parameters.
If only one number is specified, thread_pool_min and thread_pool_max are both set to this number, and thread_pool_timeout has no effect.
The following hash algorithms are available:
The following storage types are available:
Storage for each object is allocated with malloc(3).
The size parameter specifies the maximum amount of memory varnishd will allocate. The size is assumed to be in bytes, unless followed by one of the following suffixes:
K, k The size is expressed in kibibytes.
M, m The size is expressed in mebibytes.
G, g The size is expressed in gibibytes.
T, t The size is expressed in tebibytes.
The default size is unlimited.
Storage for each object is allocated from an arena backed by a file. This is the default.
The path parameter specifies either the path to the backing file or the path to a directory in which varnishd will create the backing file. The default is /tmp.
The size parameter specifies the size of the backing file. The size is assumed to be in bytes, unless fol‐ lowed by one of the following suffixes:
K, k The size is expressed in kibibytes.
M, m The size is expressed in mebibytes.
G, g The size is expressed in gibibytes.
T, t The size is expressed in tebibytes.
% The size is expressed as a percentage of the free space on the file system where it resides.
The default size is 50%.
If the backing file already exists, it will be truncated or expanded to the specified size.
Note that if varnishd has to create or expand the file, it will not pre-allocate the added space, leading to fragmentation, which may adversely impact performance. Pre-creating the storage file using dd(1) will reduce fragmentation to a minimum.
The granularity parameter specifies the granularity of allocation. All allocations are rounded up to this size. The size is assumed to be in bytes, unless followed by one of the suffixes described for size except for %.
The default size is the VM page size. The size should be reduced if you have many small objects.
If the -T option was specified, varnishd will offer a command-line management interface on the specified address and port. The following commands are available:
Display a list of available commands.
If the command is specified, display help for this command.
Display a list if run-time parameters and their values.
If the -l option is specified, the list includes a brief explanation of each parameter.
If a param is specified, display only the value and explanation for this parameter.
Display the purge list.
All requests for objects from the cache are matched against items on the purge list. If an object in the cache is older than a matching purge list item, it is considered “purged”, and will be fetched from the backend instead.
When a purge expression is older than all the objects in the cache, it is removed from the list.
Show summary statistics.
All the numbers presented are totals since server startup; for a better idea of the current situation, use the varnishstat(1) utility.
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:
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 param.show command, so should there be a new parameter which is not listed here, you can find the description using the CLI commands.
Be aware that on 32 bit systems, certain default values, such as sess_workspace (=16k) and thread_pool_stack (=64k) are reduced relative to the values listed here, in order to conserve VM space.
Default: 0.900 Flags: experimental
If we run out of resources, such as file descriptors or worker threads, the acceptor will sleep between accepts. This parameter (multiplicatively) reduce the sleep duration for each succesfull accept. (ie: 0.9 = reduce by 10%)
Units: s Default: 0.001 Flags: experimental
If we run out of resources, such as file descriptors or worker threads, the acceptor will sleep between accepts. This parameter control how much longer we sleep, each time we fail to accept a new connection.
Units: s Default: 0.050 Flags: experimental
If we run out of resources, such as file descriptors or worker threads, the acceptor will sleep between accepts. This parameter limits how long it can sleep between attempts to accept new connections.
Units: bool Default: on
Restart child process automatically if it dies.
Units: s Default: 0.0
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. A value of zero disables the ban lurker.
Units: s Default: 60
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.
Units: bool Default: off Flags: experimental
Cache vbe_conn’s or rely on malloc, that’s the question.
Default: exec cc -fpic -shared -Wl,-x -o %o %s Flags: must_reload
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.
Units: bytes Default: 8192
Size of buffer for CLI input. You may need to increase this if you have big VCL files and use the vcl.inline CLI command. NB: Must be specified with -p to have effect.
Units: seconds Default: 10
Timeout for the childs replies to CLI requests from the master.
Units: s Default: 10
How much clockskew we are willing to accept between the backend and our own clock.
Units: s Default: 0.4
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.
Default: 10seconds Flags: delayed
Default grace period. We will deliver an object this long after it has expired, provided another thread is attempting to get a new copy.
Units: seconds Default: 120
The TTL assigned to objects if neither the backend nor the VCL code assigns one. Objects already cached will not be affected by changes made until they are fetched from the backend again. To force an immediate effect at the expense of a total flush of the cache use “purge.url .”
Units: bitmap Default: 0 Bitmap controlling diagnostics code:
0x00000001 - CNT_Session states.
0x00000002 - workspace debugging.
0x00000004 - kqueue debugging.
0x00000008 - mutex logging.
0x00000010 - mutex contests.
0x00000020 - waiting list.
0x00000040 - object workspace.
0x00001000 - do not core-dump child process.
0x00002000 - only short panic message.
0x00004000 - panic to stderr.
0x00008000 - panic to abort2().
0x00010000 - synchronize shmlog.
0x00020000 - synchronous start of persistence.
0x80000000 - do edge-detection on digest.
Use 0x notation and do the bitor in your head :-)
Units: seconds Default: 0
The TTL assigned to the synthesized error pages
Units: bitmap Default: 0 Bitmap controlling ESI parsing code:
0x00000001 - Don't check if it looks like XML
0x00000002 - Ignore non-esi elements
0x00000004 - Emit parsing debug records
Use 0x notation and do the bitor in your head :-)
Units: kilobytes Default: 128 Flags: experimental
The default chunksize used by fetcher. This should be bigger than the majority of objects with short TTLs. Internal limits in the storage_file module makes increases above 128kb a dubious idea.
Units: s Default: 60
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.
Default: ..... Flags: must_restart
The unprivileged group to run as.
Units: header lines Default: 64
Maximum number of HTTP headers we will deal with. This space is preallocated in sessions and workthreads only objects allocate only space for the headers they store.
Default: off
Enables experimental support for the HTTP range header, enabling Varnish to serve parts of an object to a client. However, Varnish will request the whole object from the backend server.
Default: :80 Flags: must_restart
Whitespace separated list of network endpoints where Varnish will accept requests. Possible formats: host, host:port, :port
Units: connections Default: 1024 Flags: must_restart
Listen queue depth.
Units: bool Default: off
Log the hash string to shared memory log.
Units: bool Default: off
Log the local address on the TCP connection in the SessionOpen shared memory record.
Units: seconds Default: 2 Flags: experimental
Grace period before object moves on LRU list. 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.
Units: includes Default: 5
Maximum depth of esi:include processing.
Units: restarts Default: 4
Upper limit on how many times a request can restart. Be aware that restarts are likely to cause a hit against the backend, so don’t increase thoughtlessly.
Units: % Default: 100 Flags: experimental
Percentage permitted overflow queue length.
This sets the ratio of queued requests to worker threads, above which sessions will be dropped instead of queued.
Units: seconds Default: 3 Flags: must_restart
Interval between pings from parent to child. Zero will disable pinging entirely, which makes it possible to attach a debugger to the child.
Units: seconds Default: 60
Idle timeout for PIPE sessions. If nothing have been received in either direction for this many seconds, the session is closed.
Units: bool Default: off
Prefer IPv6 address when connecting to backends which have both IPv4 and IPv6 addresses.
Units: bool Default: on
Detect and eliminate duplicate purges.
Units: requests per request Default: 3 Flags: experimental
How many parked request we start for each completed request on the object. NB: Even with the implict delay of delivery, this parameter controls an exponential increase in number of worker threads.
Units: objects Default: 10 Flags: experimental
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.
Units: seconds Default: 600 Flags: delayed
Send timeout for client connections. If no data has been sent to the client in this many seconds, the ses‐ sion is closed. See setsockopt(2) under SO_SNDTIMEO for more information.
Units: bytes Default: -1 Flags: experimental
The minimum size of objects transmitted with sendfile.
Units: seconds Default: 5
Idle timeout for persistent sessions. If a HTTP request has not been received in this many seconds, the session is closed.
Units: bytes Default: 65536 Flags: delayed
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. Minimum is 1024 bytes.
Units: ms Default: 50 Flags: experimental
How long time the workerthread lingers on the session to see if a new request appears right away. If sessions are reused, as much as half of all reuses happen within the first 100 msec of the previous request completing. 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.
Units: sessions Default: 100000
Maximum number of sessions we will allocate before just dropping connections. This is mostly an anti-DoS measure, and setting it plenty high should not hurt, as long as you have the memory for it.
Units: bytes Default: 255
Maximum number of bytes in SHM log record. Maximum is 65535 bytes.
Units: bytes Default: 8192 Flags: delayed
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. These flushes show up in stats as “SHM flushes due to overflow”. Minimum is 4096 bytes.
Units: bool Default: on
Log all CLI traffic to syslog(LOG_INFO).
Units: milliseconds Default: 20 Flags: experimental
Wait at least this long between creating threads.
Setting this too long results in insuffient worker threads.
Setting this too short increases the risk of worker thread pile-up.
Units: requests Default: 2 Flags: experimental
Overflow threshold for worker thread creation.
Setting this too low, will result in excess worker threads, which is generally a bad idea.
Setting it too high results in insuffient worker threads.
Units: milliseconds Default: 200 Flags: experimental
Wait at least this long after a failed thread creation before trying to create another thread.
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.
If thread creation failures are a problem, check that thread_pool_max is not too high.
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.
Units: threads Default: 500 Flags: delayed, experimental
The maximum number of worker threads in all pools combined.
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.
Units: threads Default: 5 Flags: delayed, experimental
The minimum number of threads in each worker pool.
Increasing this may help ramp up faster from low load situations where threads have expired.
Minimum is 2 threads.
Units: milliseconds Default: 1000 Flags: delayed, experimental
Wait this long between purging threads.
This controls the decay of thread pools when idle(-ish).
Minimum is 100 milliseconds.
Units: bytes Default: -1 Flags: experimental
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.
Units: seconds Default: 300 Flags: delayed, experimental
Thread idle threshold.
Threads in excess of thread_pool_min, which have been idle for at least this long are candidates for purg‐ ing.
Minimum is 1 second.
Units: pools Default: 2 Flags: delayed, experimental
Number of worker thread pools.
Increasing number of worker pools decreases lock contention.
Too many pools waste CPU and RAM resources, and more than one pool for each CPU is probably detrimal to performance.
Can be increased on the fly, but decreases require a restart to take effect.
Units: requests Default: 10 Flags: experimental
Worker threads accumulate statistics, and dump these into the global stats counters if the lock is free when they finish a request. 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.
Flags: must_restart
The unprivileged user to run as. Setting this will also set “group” to the specified user’s primary group.
Units: bool Default: off
Trace VCL execution in the shmlog. Enabling this will allow you to see the path each request has taken through the VCL program. This generates a lot of logrecords so it is off by default.
Default: default Flags: must_restart, experimental
Select the waiter kernel interface.
A purge expression consists of one or more conditions. A condition consists of a field, an operator, and an argument. Conditions can be ANDed together with “&&”.
A field can be any of the variables from VCL, for instance req.url, req.http.host or obj.set-cookie.
Operators are “==” for direct comparision, “~” for a regular expression match, and “>” or “<” for size compar‐ isons. Prepending an operator with ”!” negates the expression.
The argument could be a quoted string, a regexp, or an integer. Integers can have “KB”, “MB”, “GB” or “TB” appended for size related fields.
Simple example: All requests where req.url exactly matches the string /news are purged from the cache::
req.url == "/news"
Example: Purge all documents where the name does not end with ”.ogg”, and where the size of the object is greater than 10 megabytes::
req.url !~ "\.ogg$" && obj.size > 10MB
Example: Purge all documents where the serving host is “example.com” or “www.example.com”, and where the Set- Cookie header received from the backend contains “USERID=1663”::
req.http.host ~ "^(www\.)example.com$" && obj.set-cookie ~ "USERID=1663"
The varnishd daemon was developed by Poul-Henning Kamp in cooperation with Verdens Gang AS, Linpro AS and Varnish Software.
This manual page was written by Dag-Erling Smørgrav with updates by Stig Sandbeck Mathisen ⟨ssm@debian.org⟩
This document is licensed under the same licence as Varnish itself. See LICENCE for details.