2 <!DOCTYPE modulesynopsis SYSTEM "../style/modulesynopsis.dtd">
3 <?xml-stylesheet type="text/xsl" href="../style/manual.en.xsl"?>
4 <!-- $LastChangedRevision$ -->
7 Licensed to the Apache Software Foundation (ASF) under one or more
8 contributor license agreements. See the NOTICE file distributed with
9 this work for additional information regarding copyright ownership.
10 The ASF licenses this file to You under the Apache License, Version 2.0
11 (the "License"); you may not use this file except in compliance with
12 the License. You may obtain a copy of the License at
14 http://www.apache.org/licenses/LICENSE-2.0
16 Unless required by applicable law or agreed to in writing, software
17 distributed under the License is distributed on an "AS IS" BASIS,
18 WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
19 See the License for the specific language governing permissions and
20 limitations under the License.
23 <modulesynopsis metafile="mod_proxy_ajp.xml.meta">
25 <name>mod_proxy_ajp</name>
26 <description>AJP support module for
27 <module>mod_proxy</module></description>
28 <status>Extension</status>
29 <sourcefile>mod_proxy_ajp.c</sourcefile>
30 <identifier>proxy_ajp_module</identifier>
33 <p>This module <em>requires</em> the service of <module
34 >mod_proxy</module>. It provides support for the
35 <code>Apache JServ Protocol version 1.3</code> (hereafter
38 <p>Thus, in order to get the ability of handling <code>AJP13</code>
39 protocol, <module>mod_proxy</module> and
40 <module>mod_proxy_ajp</module> have to be present in the server.</p>
42 <note type="warning"><title>Warning</title>
43 <p>Do not enable proxying until you have <a
44 href="mod_proxy.html#access">secured your server</a>. Open proxy
45 servers are dangerous both to your network and to the Internet at
50 <seealso><module>mod_proxy</module></seealso>
51 <seealso><a href="../env.html">Environment Variable documentation</a></seealso>
53 <section id="usage"><title>Usage</title>
54 <p>This module is used to reverse proxy to a backend application server
55 (e.g. Apache Tomcat) using the AJP13 protocol. The usage is similar to
56 an HTTP reverse proxy, but uses the <code>ajp://</code> prefix:</p>
58 <example><title>Simple Reverse Proxy</title>
59 <highlight language="config">
60 ProxyPass /app ajp://backend.example.com:8009/app
64 <p>Balancers may also be used:</p>
65 <example><title>Balancer Reverse Proxy</title>
66 <highlight language="config">
67 <Proxy balancer://cluster>
68 BalancerMember ajp://app1.example.com:8009 loadfactor=1
69 BalancerMember ajp://app2.example.com:8009 loadfactor=2
70 ProxySet lbmethod=bytraffic
72 ProxyPass /app balancer://cluster/app
76 <p>Note that usually no
77 <directive module="mod_proxy">ProxyPassReverse</directive>
78 directive is necessary. The AJP request includes the original host
79 header given to the proxy, and the application server can be expected
80 to generate self-referential headers relative to this host, so no
81 rewriting is necessary.</p>
83 <p>The main exception is when the URL path on the proxy differs from that
85 backend. In this case, a redirect header can be rewritten relative to the
86 original host URL (not the backend <code>ajp://</code> URL), for
88 <example><title>Rewriting Proxied Path</title>
89 <highlight language="config">
90 ProxyPass /apps/foo ajp://backend.example.com:8009/foo
91 ProxyPassReverse /apps/foo http://www.example.com/foo
94 <p>However, it is usually better to deploy the application on the backend
95 server at the same path as the proxy rather than to take this approach.
99 <section id="env"><title>Environment Variables</title>
100 <p>Environment variables whose names have the prefix <code>AJP_</code>
101 are forwarded to the origin server as AJP request attributes
102 (with the AJP_ prefix removed from the name of the key).</p>
105 <section id="overviewprotocol"><title>Overview of the protocol</title>
106 <p>The <code>AJP13</code> protocol is packet-oriented. A binary format
107 was presumably chosen over the more readable plain text for reasons of
108 performance. The web server communicates with the servlet container over
109 TCP connections. To cut down on the expensive process of socket creation,
110 the web server will attempt to maintain persistent TCP connections to the
111 servlet container, and to reuse a connection for multiple request/response
113 <p>Once a connection is assigned to a particular request, it will not be
114 used for any others until the request-handling cycle has terminated. In
115 other words, requests are not multiplexed over connections. This makes
116 for much simpler code at either end of the connection, although it does
117 cause more connections to be open at once.</p>
118 <p>Once the web server has opened a connection to the servlet container,
119 the connection can be in one of the following states:</p>
121 <li> Idle <br/> No request is being handled over this connection. </li>
122 <li> Assigned <br/> The connection is handling a specific request.</li>
124 <p>Once a connection is assigned to handle a particular request, the basic
125 request information (e.g. HTTP headers, etc) is sent over the connection in
126 a highly condensed form (e.g. common strings are encoded as integers).
127 Details of that format are below in Request Packet Structure. If there is a
128 body to the request <code>(content-length > 0)</code>, that is sent in a
129 separate packet immediately after.</p>
130 <p>At this point, the servlet container is presumably ready to start
131 processing the request. As it does so, it can send the
132 following messages back to the web server:</p>
134 <li>SEND_HEADERS <br/>Send a set of headers back to the browser.</li>
135 <li>SEND_BODY_CHUNK <br/>Send a chunk of body data back to the browser.
137 <li>GET_BODY_CHUNK <br/>Get further data from the request if it hasn't all
138 been transferred yet. This is necessary because the packets have a fixed
139 maximum size and arbitrary amounts of data can be included the body of a
140 request (for uploaded files, for example). (Note: this is unrelated to
141 HTTP chunked transfer).</li>
142 <li>END_RESPONSE <br/> Finish the request-handling cycle.</li>
144 <p>Each message is accompanied by a differently formatted packet of data.
145 See Response Packet Structures below for details.</p>
148 <section id="basppacketstruct"><title>Basic Packet Structure</title>
149 <p>There is a bit of an XDR heritage to this protocol, but it differs
150 in lots of ways (no 4 byte alignment, for example).</p>
151 <p>Byte order: I am not clear about the endian-ness of the individual
152 bytes. I'm guessing the bytes are little-endian, because that's what
153 XDR specifies, and I'm guessing that sys/socket library is magically
154 making that so (on the C side). If anyone with a better knowledge of
155 socket calls can step in, that would be great.</p>
156 <p>There are four data types in the protocol: bytes, booleans,
157 integers and strings.</p>
159 <dt><strong>Byte</strong></dt><dd>A single byte.</dd>
160 <dt><strong>Boolean</strong></dt>
161 <dd>A single byte, <code>1 = true</code>, <code>0 = false</code>.
162 Using other non-zero values as true (i.e. C-style) may work in some places,
163 but it won't in others.</dd>
164 <dt><strong>Integer</strong></dt>
165 <dd>A number in the range of <code>0 to 2^16 (32768)</code>. Stored in
166 2 bytes with the high-order byte first.</dd>
167 <dt><strong>String</strong></dt>
168 <dd>A variable-sized string (length bounded by 2^16). Encoded with
169 the length packed into two bytes first, followed by the string
170 (including the terminating '\0'). Note that the encoded length does
171 <strong>not</strong> include the trailing '\0' -- it is like
172 <code>strlen</code>. This is a touch confusing on the Java side, which
173 is littered with odd autoincrement statements to skip over these
174 terminators. I believe the reason this was done was to allow the C
175 code to be extra efficient when reading strings which the servlet
176 container is sending back -- with the terminating \0 character, the
177 C code can pass around references into a single buffer, without copying.
178 if the \0 was missing, the C code would have to copy things out in order
179 to get its notion of a string.</dd>
182 <section><title>Packet Size</title>
183 <p>According to much of the code, the max packet size is <code>
184 8 * 1024 bytes (8K)</code>. The actual length of the packet is encoded in
187 <section><title>Packet Headers</title>
188 <p>Packets sent from the server to the container begin with
189 <code>0x1234</code>. Packets sent from the container to the server
190 begin with <code>AB</code> (that's the ASCII code for A followed by the
191 ASCII code for B). After those first two bytes, there is an integer
192 (encoded as above) with the length of the payload. Although this might
193 suggest that the maximum payload could be as large as 2^16, in fact, the
194 code sets the maximum to be 8K.</p>
197 <td colspan="6"><em>Packet Format (Server->Container)</em></td>
211 <td colspan="2">Data Length (n)</td>
217 <td colspan="6"><em>Packet Format (Container->Server)</em></td>
231 <td colspan="2">Data Length (n)</td>
235 <p>For most packets, the first byte of the payload encodes the type of
236 message. The exception is for request body packets sent from the server to
237 the container -- they are sent with a standard packet header (<code>
238 0x1234</code> and then length of the packet), but without any prefix code
240 <p>The web server can send the following messages to the servlet
245 <td>Type of Packet</td>
250 <td>Forward Request</td>
251 <td>Begin the request-processing cycle with the following data</td>
256 <td>The web server asks the container to shut itself down.</td>
261 <td>The web server asks the container to take control
262 (secure login phase).</td>
267 <td>The web server asks the container to respond quickly with a CPong.
273 <td>Size (2 bytes) and corresponding body data.</td>
276 <p>To ensure some basic security, the container will only actually do the
277 <code>Shutdown</code> if the request comes from the same machine on which
279 <p>The first <code>Data</code> packet is send immediately after the
280 <code>Forward Request</code> by the web server.</p>
281 <p>The servlet container can send the following types of messages to the
286 <td>Type of Packet</td>
291 <td>Send Body Chunk</td>
292 <td>Send a chunk of the body from the servlet container to the web
293 server (and presumably, onto the browser). </td>
297 <td>Send Headers</td>
298 <td>Send the response headers from the servlet container to the web
299 server (and presumably, onto the browser).</td>
303 <td>End Response</td>
304 <td>Marks the end of the response (and thus the request-handling cycle).
309 <td>Get Body Chunk</td>
310 <td>Get further data from the request if it hasn't all been
311 transferred yet.</td>
316 <td>The reply to a CPing request</td>
319 <p>Each of the above messages has a different internal structure, detailed
323 <section id="rpacetstruct"><title>Request Packet Structure</title>
324 <p>For messages from the server to the container of type
325 <em>Forward Request</em>:</p>
327 AJP13_FORWARD_REQUEST :=
328 prefix_code (byte) 0x02 = JK_AJP13_FORWARD_REQUEST
335 server_port (integer)
337 num_headers (integer)
338 request_headers *(req_header_name req_header_value)
339 attributes *(attribut_name attribute_value)
340 request_terminator (byte) OxFF
342 <p>The <code>request_headers</code> have the following structure:
345 sc_req_header_name | (string) [see below for how this is parsed]
347 sc_req_header_name := 0xA0xx (integer)
349 req_header_value := (string)
351 <p>The <code>attributes</code> are optional and have the following
354 attribute_name := sc_a_name | (sc_a_req_attribute string)
356 attribute_value := (string)
359 <p>Not that the all-important header is <code>content-length</code>,
360 because it determines whether or not the container looks for another
361 packet immediately.</p>
362 <section><title>Detailed description of the elements of Forward Request
364 <section><title>Request prefix</title>
365 <p>For all requests, this will be 2. See above for details on other Prefix
368 <section><title>Method</title>
369 <p>The HTTP method, encoded as a single byte:</p>
371 <tr><td>Command Name</td><td>Code</td></tr>
372 <tr><td>OPTIONS</td><td>1</td></tr>
373 <tr><td>GET</td><td>2</td></tr>
374 <tr><td>HEAD</td><td>3</td></tr>
375 <tr><td>POST</td><td>4</td></tr>
376 <tr><td>PUT</td><td>5</td></tr>
377 <tr><td>DELETE</td><td>6</td></tr>
378 <tr><td>TRACE</td><td>7</td></tr>
379 <tr><td>PROPFIND</td><td>8</td></tr>
380 <tr><td>PROPPATCH</td><td>9</td></tr>
381 <tr><td>MKCOL</td><td>10</td></tr>
382 <tr><td>COPY</td><td>11</td></tr>
383 <tr><td>MOVE</td><td>12</td></tr>
384 <tr><td>LOCK</td><td>13</td></tr>
385 <tr><td>UNLOCK</td><td>14</td></tr>
386 <tr><td>ACL</td><td>15</td></tr>
387 <tr><td>REPORT</td><td>16</td></tr>
388 <tr><td>VERSION-CONTROL</td><td>17</td></tr>
389 <tr><td>CHECKIN</td><td>18</td></tr>
390 <tr><td>CHECKOUT</td><td>19</td></tr>
391 <tr><td>UNCHECKOUT</td><td>20</td></tr>
392 <tr><td>SEARCH</td><td>21</td></tr>
393 <tr><td>MKWORKSPACE</td><td>22</td></tr>
394 <tr><td>UPDATE</td><td>23</td></tr>
395 <tr><td>LABEL</td><td>24</td></tr>
396 <tr><td>MERGE</td><td>25</td></tr>
397 <tr><td>BASELINE_CONTROL</td><td>26</td></tr>
398 <tr><td>MKACTIVITY</td><td>27</td></tr>
400 <p>Later version of ajp13, will transport
401 additional methods, even if they are not in this list.</p>
403 <section><title>protocol, req_uri, remote_addr, remote_host, server_name,
404 server_port, is_ssl</title>
405 <p>These are all fairly self-explanatory. Each of these is required, and
406 will be sent for every request.</p>
408 <section><title>Headers</title>
409 <p>The structure of <code>request_headers</code> is the following:
410 First, the number of headers <code>num_headers</code> is encoded.
411 Then, a series of header name <code>req_header_name</code> / value
412 <code>req_header_value</code> pairs follows.
413 Common header names are encoded as integers,
414 to save space. If the header name is not in the list of basic headers,
415 it is encoded normally (as a string, with prefixed length). The list of
416 common headers <code>sc_req_header_name</code>and their codes
417 is as follows (all are case-sensitive):</p>
419 <tr><td>Name</td><td>Code value</td><td>Code name</td></tr>
420 <tr><td>accept</td><td>0xA001</td><td>SC_REQ_ACCEPT</td></tr>
421 <tr><td>accept-charset</td><td>0xA002</td><td>SC_REQ_ACCEPT_CHARSET
423 <tr><td>accept-encoding</td><td>0xA003</td><td>SC_REQ_ACCEPT_ENCODING
425 <tr><td>accept-language</td><td>0xA004</td><td>SC_REQ_ACCEPT_LANGUAGE
427 <tr><td>authorization</td><td>0xA005</td><td>SC_REQ_AUTHORIZATION</td>
429 <tr><td>connection</td><td>0xA006</td><td>SC_REQ_CONNECTION</td></tr>
430 <tr><td>content-type</td><td>0xA007</td><td>SC_REQ_CONTENT_TYPE</td>
432 <tr><td>content-length</td><td>0xA008</td><td>SC_REQ_CONTENT_LENGTH</td>
434 <tr><td>cookie</td><td>0xA009</td><td>SC_REQ_COOKIE</td></tr>
435 <tr><td>cookie2</td><td>0xA00A</td><td>SC_REQ_COOKIE2</td></tr>
436 <tr><td>host</td><td>0xA00B</td><td>SC_REQ_HOST</td></tr>
437 <tr><td>pragma</td><td>0xA00C</td><td>SC_REQ_PRAGMA</td></tr>
438 <tr><td>referer</td><td>0xA00D</td><td>SC_REQ_REFERER</td></tr>
439 <tr><td>user-agent</td><td>0xA00E</td><td>SC_REQ_USER_AGENT</td></tr>
441 <p>The Java code that reads this grabs the first two-byte integer and if
442 it sees an <code>'0xA0'</code> in the most significant
443 byte, it uses the integer in the second byte as an index into an array of
444 header names. If the first byte is not <code>0xA0</code>, it assumes that
445 the two-byte integer is the length of a string, which is then read in.</p>
446 <p>This works on the assumption that no header names will have length
447 greater than <code>0x9999 (==0xA000 - 1)</code>, which is perfectly
448 reasonable, though somewhat arbitrary.</p>
449 <note><title>Note:</title>
450 The <code>content-length</code> header is extremely
451 important. If it is present and non-zero, the container assumes that
452 the request has a body (a POST request, for example), and immediately
453 reads a separate packet off the input stream to get that body.
456 <section><title>Attributes</title>
457 <p>The attributes prefixed with a <code>?</code>
458 (e.g. <code>?context</code>) are all optional. For each, there is a
459 single byte code to indicate the type of attribute, and then its value
460 (string or integer). They can be sent in any order (though the C code
461 always sends them in the order listed below). A special terminating code
462 is sent to signal the end of the list of optional attributes. The list of
465 <tr><td>Information</td><td>Code Value</td><td>Type Of Value</td><td>Note</td></tr>
466 <tr><td>?context</td><td>0x01</td><td>-</td><td>Not currently implemented
468 <tr><td>?servlet_path</td><td>0x02</td><td>-</td><td>Not currently implemented
470 <tr><td>?remote_user</td><td>0x03</td><td>String</td><td></td></tr>
471 <tr><td>?auth_type</td><td>0x04</td><td>String</td><td></td></tr>
472 <tr><td>?query_string</td><td>0x05</td><td>String</td><td></td></tr>
473 <tr><td>?jvm_route</td><td>0x06</td><td>String</td><td></td></tr>
474 <tr><td>?ssl_cert</td><td>0x07</td><td>String</td><td></td></tr>
475 <tr><td>?ssl_cipher</td><td>0x08</td><td>String</td><td></td></tr>
476 <tr><td>?ssl_session</td><td>0x09</td><td>String</td><td></td></tr>
477 <tr><td>?req_attribute</td><td>0x0A</td><td>String</td><td>Name (the name of the
478 attribute follows)</td></tr>
479 <tr><td>?ssl_key_size</td><td>0x0B</td><td>Integer</td><td></td></tr>
480 <tr><td>are_done</td><td>0xFF</td><td>-</td><td>request_terminator</td></tr>
482 <p>The <code>context</code> and <code>servlet_path</code> are not
483 currently set by the C code, and most of the Java code completely ignores
484 whatever is sent over for those fields (and some of it will actually break
485 if a string is sent along after one of those codes). I don't know if this
486 is a bug or an unimplemented feature or just vestigial code, but it's
487 missing from both sides of the connection.</p>
488 <p>The <code>remote_user</code> and <code>auth_type</code> presumably
489 refer to HTTP-level authentication, and communicate the remote user's
490 username and the type of authentication used to establish their identity
491 (e.g. Basic, Digest).</p>
492 <p>The <code>query_string</code>, <code>ssl_cert</code>,
493 <code>ssl_cipher</code>, and <code>ssl_session</code> refer to the
494 corresponding pieces of HTTP and HTTPS.</p>
495 <p>The <code>jvm_route</code>, is used to support sticky
496 sessions -- associating a user's sesson with a particular Tomcat instance
497 in the presence of multiple, load-balancing servers.</p>
498 <p>Beyond this list of basic attributes, any number of other attributes
499 can be sent via the <code>req_attribute</code> code <code>0x0A</code>.
500 A pair of strings to represent the attribute name and value are sent
501 immediately after each instance of that code. Environment values are passed
502 in via this method.</p>
503 <p>Finally, after all the attributes have been sent, the attribute
504 terminator, <code>0xFF</code>, is sent. This signals both the end of the
505 list of attributes and also then end of the Request Packet.</p>
509 <section id="resppacketstruct"><title>Response Packet Structure</title>
510 <p>for messages which the container can send back to the server.</p>
512 AJP13_SEND_BODY_CHUNK :=
514 chunk_length (integer)
516 chunk_terminator (byte) Ox00
519 AJP13_SEND_HEADERS :=
521 http_status_code (integer)
522 http_status_msg (string)
523 num_headers (integer)
524 response_headers *(res_header_name header_value)
527 sc_res_header_name | (string) [see below for how this is parsed]
529 sc_res_header_name := 0xA0 (byte)
531 header_value := (string)
533 AJP13_END_RESPONSE :=
538 AJP13_GET_BODY_CHUNK :=
540 requested_length (integer)
542 <section><title>Details:</title></section>
543 <section><title>Send Body Chunk</title>
544 <p>The chunk is basically binary data, and is sent directly back to the
547 <section><title>Send Headers</title>
548 <p>The status code and message are the usual HTTP things
549 (e.g. <code>200</code> and <code>OK</code>). The response header names are
550 encoded the same way the request header names are. See header_encoding above
551 for details about how the codes are distinguished from the strings.<br />
552 The codes for common headers are:</p>
554 <tr><td>Name</td><td>Code value</td></tr>
555 <tr><td>Content-Type</td><td>0xA001</td></tr>
556 <tr><td>Content-Language</td><td>0xA002</td></tr>
557 <tr><td>Content-Length</td><td>0xA003</td></tr>
558 <tr><td>Date</td><td>0xA004</td></tr>
559 <tr><td>Last-Modified</td><td>0xA005</td></tr>
560 <tr><td>Location</td><td>0xA006</td></tr>
561 <tr><td>Set-Cookie</td><td>0xA007</td></tr>
562 <tr><td>Set-Cookie2</td><td>0xA008</td></tr>
563 <tr><td>Servlet-Engine</td><td>0xA009</td></tr>
564 <tr><td>Status</td><td>0xA00A</td></tr>
565 <tr><td>WWW-Authenticate</td><td>0xA00B</td></tr>
567 <p> After the code or the string header name, the header value is
568 immediately encoded.</p>
570 <section><title>End Response</title>
571 <p>Signals the end of this request-handling cycle. If the
572 <code>reuse</code> flag is true <code>(==1)</code>, this TCP connection can
573 now be used to handle new incoming requests. If <code>reuse</code> is false
574 (anything other than 1 in the actual C code), the connection should
577 <section><title>Get Body Chunk</title>
578 <p>The container asks for more data from the request (If the body was
579 too large to fit in the first packet sent over or when the request is
580 chunked). The server will send a body packet back with an amount of data
581 which is the minimum of the <code>request_length</code>, the maximum send
582 body size <code>(8186 (8 Kbytes - 6))</code>, and the number of bytes
583 actually left to send from the request body.<br/>
584 If there is no more data in the body (i.e. the servlet container is
585 trying to read past the end of the body), the server will send back an
586 <em>empty</em> packet, which is a body packet with a payload length of 0.
587 <code>(0x12,0x34,0x00,0x00)</code></p>