purge
=====

The purge tool is a kind of magnifying glass into your squid-2 cache. You
can use purge to have a look at what URLs are stored in which file within
your cache. The purge tool can also be used to release objects which URLs
match user specified regular expressions. A more troublesome feature is the
ability to remove files squid does not seem to know about any longer.

    USE AT YOUR OWN RISK! NO GUARANTEES, WHATSOEVER! DON'T BLAME US!
			 YOU HAVE BEEN WARNED!



compilation
===========

Purge has been successfully compiled under the following OSes:

	SYSTEM		g++	native
	------		---	------
	Solaris 2.7	yes	CC
	IRIX 6.5	yes	CC -n32
	Linux 2.0.36	yes	(g++ IS native)
	FreeBSD 4.x	yes	gmake port must be installed
				(g++ IS supported)

The recent move of the Linux community towards glibc2 may cause some
troubles, though. The compilation requires GNU make, no other make will work
correctly. The source distribution contains all files checked into the
revision control repository. Therefore, you will need to install GNU RCS
first (which in turn needs the GNU diffutils). 

The repository also contains the prototypical Perl implementation. The user
interface in the C++ implementation changed a little when compared to the
Perl one. You will have to state at least one regular expression for purge
to start working. Also, printing the complete cache URLs, you will need to
specify the "-e ." regular expression.

In order to compile the purge tool, untar the source distribution and
change into the purge directory. With RCS and GNU make installed, just say
"make". GNU make will automagically retrieve all necessary files from the
repository and create the binary. 

Systems not stated above will need to retrieve the makefile (use "co -l
Makefile" for this) and add their own platform specific definitions to
section [2] in the makefile.



squid preparation
=================

In order to use purge for real PURGEs, you will have to enable this feature
in squid. By default, PURGE is disabled. You should watch closely for whom
you enable the PURGE ability, otherwise total stranger just might wipe your
cache content. The following lines will need to be added to your squid.conf
(you may want to add further networks to the src_local ACL):

	acl purge method PURGE
	acl src_local src 127.0.0.0/8
	http_access allow purge src_local
	http_access deny  purge

Reconfigure or restart (preferred) your squid after changing the
configuration file.



modes of operation
==================

$Id: purge.cc,v 1.15 2000/09/21 09:05:56 cached Exp $
Usage:  purge   [-a] [-c cf] [-d l] [-(f|F) fn | -(e|E) re] [-p h[:p]]
                [-P #] [-s] [-v] [-C dir [-H]] [-n]

 -a     display a little rotating thingy to indicate that I am alive (tty only).
 -c c   squid.conf location, default "/usr/local/squid/etc/squid.conf".
 -C dir base directory for content extraction (copy-out mode).
 -d l   debug level, an OR of different debug options.
 -e re  single regular expression per -e instance (use quotes!).
 -E re  single case sensitive regular expression like -e.
 -f fn  name of textfile containing one regular expression per line.
 -F fn  name of textfile like -f containing case sensitive REs.
 -H     prepend HTTP reply header to destination files in copy-out mode.
 -n     do not fork() when using more than one cache_dir.
 -p h:p cache runs on host h and optional port p, default is localhost:3128.
 -P #   if 0, just print matches; otherwise OR the following purge modes:
           0x01 really send PURGE to the cache.
           0x02 remove all caches files reported as 404 (not found).
           0x04 remove all weird (inaccessible or too small) cache files.
        0 and 1 are recommended - slow rebuild your cache with other modes.
 -s     show all options after option parsing, but before really starting.
 -v     show more information about the file, e.g. MD5, timestamps and flags.

--- &< snip, snip ---

-a	is a kind of "i am alive" flag. It can only be activated, if
	your stdout is a tty. If active, it will display a little 
	rotating line to indicate that there is actually something
	happening. You should not use this switch, if you capture 
	your stdout in a file, or if your expression list produces
	many matches. The -a flag is also incompatible with the
	(default) multi cache_dir mode.

	default: off
	See also: -n

-c cd	CHANGED!
	this option lets you specify the location of the squid.conf file.
	Purge now understands about more than one cache_dir, and does so
	by parsing Squid's configuration file. It knows about both ways
	of Squid-2 cache_dir specifications, and will automatically try
	to use the correct one.

	default: /usr/local/squid/etc/squid.conf

-C cd	if you want to rescue files from your cache, you need to specify
	the directory into which the files will be copied. Please note
	that purge will try to establish the original server's directory 
	structure. This switch also activates copy-out mode. Please do 
	not use copy-out mode with any purge mode (-P) other than 0.

	For instance, if you specified "-C /tmp", Purge will try to 
	recreate /tmp/www.server.1/url/path/file, and so forth. 

	default: off
	See also: -H, -P

-d l	lets you specify a debug level. Differents bits are reserved for
	different output.

	default: 0

-e re	the "-e" options let you specify one regular expression at the 
-E re	commandline. This is useful, if there is only a handful you
	want to check. Please remember to escape the shell metachars
	used in your regular expression. The use of single quotes 
	around your expression is recommended. The capital letter
	version works case sensitive, the lower caps version does not.

	default: (no default)

-f fn	if you have more than a handful of expression, or want to check
-F fn	the same set at regular intervals, the file option might be more
	useful to you. Each line in the text file will be regarded as
	one regular expression.  Again, the capital letter version works
	case sensitive, the lower caps version does not.

	default: (no default)

-H	if in copy-out mode (see: -C), you can specify to keep the
	HTTP Header in the recreated file. 

	default: off
	See also: -C

-n	by specifying the "-n" switch, you will tell Purge to process 
	one cache_dir after another, instead of doing things in parallel.
	If you have more than one cache_dir in your configuration,
	Purge will fork off a worker process for each cache_dir to
	do the checks for optimum speed - assuming a decently designed
	cache. Since parallel execution will put quite some load on the 
	system and its controllers, it is sometimes preferred to use 
	less resources,	though it will take longer. 

	default: parallel mode for more than one cache_dir

-p h[:p] Some cache admins (i.e. me) use a different port than 3128. The
	purge tool will need to connect to your cache in order to send
	the PURGE request (see -P). This option lets you specify the
	host and port to connect to. The port is optional. The port
	can be a name (check your /etc/services) or number. It is
	separated from the host name portion by a single colon, no
	spaces allowed.

	default: localhost:3128

-P #	If you want to do more than just print your cache content, you
	will need to specify this option. Each bit is reserved for a
	different action. Only the use of the LSB is recommended, the
	rest should be considered experimental.

		no bit set:	just print
		bit#0 set:	send PURGE for matches
		bit#1 set:	unlink object file for 404 not found PURGEs
		bit#2 set:	unlink weird object files

	If you use a value other than 0 or 1, you will need to slow
	rebuild your cache content. A warning message will remind you
	of that. If you use bit#1, all unsuccessful PURGEs will result
	in the object file in your cache directory to be removed, because
	squid does not seem to know about it any longer. Beware that the
	asyncio might try to remove it after the purge tool, and thus
	complains bitterly. Bit#1 only makes sense, if Bit#0 is also
	set, otherwise it has no effect (since the HTTP status 404 is
	never returned).

	Bit#2 is reserved for strange files which do not even contain
	a URL. Beware that these files may indicate a new object squid
	currently intends to swap onto disk. If the file suddenly went
	away, or is removed when squid tries to fetch the object, it
	will complain bitterly. You must slow rebuild your cache, if
	you use this option.

	It is recommended that if you dare to use bit#1 or bit#2, you
	should only grant the purge tool access to your squid, e.g. 
	move the HTTP and ICP listening port of squid to a different
	non-standard location during the purge.

	default: 0 (just print)

-s	If you specify this switch, all commandline parameters will be
	shown after they were parsed.

	default: off

-v	be verbose in the things reported about the file. See the output
	section below.


output
======

In regular mode, the output of purge consists of four columns. If the
URL contains not encoded whitespaces, it may look as if there are more
columns, but the last one is the URI.

 # name   meaning
 - ------ -----------------------------------------------------------
 1 file   name of cache file eximed which matches the re.
 2 status return result of purge request, "  0" in print mode.
 3 size   object size including stored headers, not file size.
 4 uri    perceived uri

Example for non-verbose output in print-mode:

/cache3/00/00/0000004A   0     5682 http://graphics.userfriendly.org/images/slovenia.gif

In verbose mode, additional columns are inserted before the uri. Time
stamps are reported using hexadecimal notation, and Squid's standard
for reporting "no such timestamp" == -1, and "unparsable timestamp" == -2.

 # name   meaning
 - ------ -----------------------------------------------------------
 1 file   name of cache file eximed which matches the re.
 2 status return result of purge request, "  0" in print mode "-P 0".
 3 size   object size including stored headers, not file size.
 4 md5    MD5 of URI from file, or "(no_md5_data_available)" string.
 5 ts     UTC of Value of Date: header in hex notation
 6 lr     UTC of last time the object was referenced
 7 ex     UTC of Expires: header
 8 lr     UTC of Last-Modified: header
 9 flags  Value of objects flags field in hex, see: Programmers Guide
10 refcnt number of times the object was referenced.
11 uri    STORE_META_URL uri or "strange_file"

Example for verbose output in print-mode:

/cache1/00/00/000000B7   0      406 7CFCB1D319F158ADC9CFD991BB8F6DCE 397d449b 39bf677b ffffffff 3820abfc 0460     1  http://www.netscape.com/images/nc_vera_tile.gif


hexd
====

The hexd tool let's you conveniently hex dump a file both, in hex char and
display char columns. Hexd only assumes that characters 0-31,127-159,255
are not printable.


$ ./hexd /cache1/00/00/000000B7 | less -r

00000000: 03 00 00 00 6D 03 00 00-00 10 7C FC B1 D3 19 F1  ....m.....|ü±Ó.ñ
00000010: 58 AD C9 CF D9 91 BB 8F-6D CE 05 00 00 00 18 39  X­ÉÏÙ.».mÎ.....9
00000020: 7D 44 9B 39 BF 67 7B FF-FF FF FF 38 20 AB FC 00  }D.9¿g{....8 «ü.
00000030: 00 00 00 00 01 04 60 04-00 00 00 30 68 74 74 70  ......`....0http
00000040: 3A 2F 2F 77 77 77 2E 6E-65 74 73 63 61 70 65 2E  ://www.netscape.
00000050: 63 6F 6D 2F 69 6D 61 67-65 73 2F 6E 63 5F 76 65  com/images/nc_ve
00000060: 72 61 5F 74 69 6C 65 2E-67 69 66 00 08 48 54 54  ra_tile.gif..HTT
00000070: 50 2F 31 2E 30 20 32 30-30 20 4F 4B 0D 0A 53 65  P/1.0 200 OK..Se
00000080: 72 76 65 72 3A 20 4E 65-74 73 63 61 70 65 2D 45  rver: Netscape-E
00000090: 6E 74 65 72 70 72 69 73-65 2F 33 2E 36 0D 0A 44  nterprise/3.6..D
000000A0: 61 74 65 3A 20 54 75 65-2C 20 32 35 20 4A 75 6C  ate: Tue, 25 Jul
000000B0: 20 32 30 30 30 20 30 37-3A 34 31 3A 31 35 20 47   2000 07:41:15 G
000000C0: 4D 54 0D 0A 43 6F 6E 74-65 6E 74 2D 54 79 70 65  MT..Content-Type
000000D0: 3A 20 69 6D 61 67 65 2F-67 69 66 0D 0A 4C 61 73  : image/gif..Las
000000E0: 74 2D 4D 6F 64 69 66 69-65 64 3A 20 57 65 64 2C  t-Modified: Wed,
000000F0: 20 30 33 20 4E 6F 76 20-31 39 39 39 20 32 31 3A   03 Nov 1999 21:
00000100: 34 31 3A 31 36 20 47 4D-54 0D 0A 43 6F 6E 74 65  41:16 GMT..Conte
00000110: 6E 74 2D 4C 65 6E 67 74-68 3A 20 36 37 0D 0A 41  nt-Length: 67..A
00000120: 63 63 65 70 74 2D 52 61-6E 67 65 73 3A 20 62 79  ccept-Ranges: by
00000130: 74 65 73 0D 0A 41 67 65-3A 20 31 38 32 37 31 33  tes..Age: 182713
00000140: 0D 0A 58 2D 43 61 63 68-65 3A 20 48 49 54 20 66  ..X-Cache: HIT f
00000150: 72 6F 6D 20 63 73 2D 68-61 6E 34 2E 77 69 6E 2D  rom cs-han4.win-
00000160: 69 70 2E 64 66 6E 2E 64-65 0D 0A 58 2D 43 61 63  ip.dfn.de..X-Cac
00000170: 68 65 2D 4C 6F 6F 6B 75-70 3A 20 48 49 54 20 66  he-Lookup: HIT f
00000180: 72 6F 6D 20 63 73 2D 68-61 6E 34 2E 77 69 6E 2D  rom cs-han4.win-
00000190: 69 70 2E 64 66 6E 2E 64-65 3A 38 30 38 31 0D 0A  ip.dfn.de:8081..
000001A0: 50 72 6F 78 79 2D 43 6F-6E 6E 65 63 74 69 6F 6E  Proxy-Connection
000001B0: 3A 20 6B 65 65 70 2D 61-6C 69 76 65 0D 0A 0D 0A  : keep-alive....
000001C0: 47 49 46 38 39 61 01 00-26 00 A2 00 00 00 00 00  GIF89a..&.¢.....
000001D0: FF FF FF 00 33 66 33 66-99 FF FF FF 00 00 00 00  ....3f3f........
000001E0: 00 00 00 00 00 21 F9 04-01 00 00 04 00 2C 00 00  .....!ù......,..
000001F0: 00 00 01 00 26 00 00 03-08 38 A2 BC DE F0 C9 A8  ....&....8¢¼Þðɨ
00000200: 12 00 3B                                         ..;



limitations
===========

o Purge does not slow rebuild the cache for you.

o It is still relatively slow, especially if your machine is low on memory
and/or unable to hold all OS directory cache entries in main memory.

o should never be used on "busy" caches with purge modes higher than 1.


TODO
====

1) use the stat() result on weird files to have a look at their ctime and 
   mtime. If they are younger than, lets say 30 seconds, they were just
   created by squid and should not be removed.

2) Add a query before purging objects or removing files, and add another
   option to remove nagging for the experienced user.

3) The reported object size may be off by one.