Issue Details (XML | Word | Printable)

Key: HUDSON-6738
Type: Bug Bug
Status: Resolved Resolved
Resolution: Fixed
Priority: Minor Minor
Assignee: kohsuke
Reporter: pzz
Votes: 2
Watchers: 5
Operations

If you were logged in you would be able to see more operations.
Hudson

Hudson doesn't start on tomcat 7

Created: 10/Jun/10 04:18 AM   Updated: Friday 07:11 AM   Resolved: 21/Jul/10 08:14 PM
Component/s: deploy
Affects Version/s: current
Fix Version/s: None

Environment: PLD Linux th, tomcat-7.0.0-rc4
Issue Links:
Dependency


 Description  « Hide

I'm trying to deploy hudson 1.361 into tomcat 7.0.0 rc4. (In fact I had working installation of hudson in tomcat 6.0.26 and I just upgraded tomcat, but I have same issue on fresh install).

Tomcat says: 'OK - Started application at context path /hudson, but when I navigate to /hudson/, I'm getting e404', but when I navigate to /hudson/, I get HTTP 404 NOT FOUND error.
There are no errors exceptions in catalina log. I tried on java-sun-1.6.0.20 and Icedtea6, same problem on both.

My context.xml is:

# cat /etc/tomcat/Catalina/localhost/hudson.xml                                                     
<?xml version="1.0" encoding="UTF-8"?>
<!-- $Id: hudson-context.xml,v 1.2 2009/02/01 23:26:40 pawelz Exp $ -->
<Context path="/hudson" docBase="/usr/share/hudson"
        privileged="false" allowLinking="true">

</Context>

(same context works perfectly with tomcat-6.0.26)

Please let me know if I can provide you with more details/information. I have testing environment, where I can test patches/workarounds/etc.

PS.: I'm setting "Priority" to minor, because tomcat-7.0.0-rc4 is not released yet, and after downgrade to tomcat-6.0.26 everything works.
PS2.: I'm not 1000% sure, that my tomcat is configured correctly, I'm just experimenting with tomcat 7 branch. But other applications (like for example Atlassian JIRA) works.



Sort Order: Ascending order - Click to sort in descending order
jieryn added a comment - 10/Jun/10 06:21 AM

Please set all Tomcat logging to FINEST.

http://tomcat.apache.org/tomcat-6.0-doc/logging.html

I don't think I've read anything about Tomcat 7 changing the way it is configured for logging so the 6.x docs should work.


pzz added a comment - 10/Jun/10 11:15 AM

One new important observation: In fact hudson starts. For example it creates files and directories in its homedir. I'm just unable to open its web interface (HTTP 404 NOT FOUND).

Jieryn, thank you for advice. I'v set log level to ALL, there are tomcat logs:

catalina.out
hudson home directory: /var/lib/hudson
catalina.err
Jun 10, 2010 6:54:51 PM org.apache.catalina.core.AprLifecycleListener init
INFO: Loaded APR based Apache Tomcat Native library 1.1.20.
Jun 10, 2010 6:54:51 PM org.apache.catalina.core.AprLifecycleListener init
INFO: APR capabilities: IPv6 [true], sendfile [true], accept filters [false], random [true].
Jun 10, 2010 6:54:51 PM org.apache.catalina.startup.Catalina load
INFO: Initialization processed in 627 ms
Jun 10, 2010 6:54:51 PM org.apache.catalina.core.StandardService startInternal
INFO: Starting service Catalina
Jun 10, 2010 6:54:51 PM org.apache.catalina.core.StandardEngine startInternal
INFO: Starting Servlet Engine: Apache Tomcat/7.0.0-RC4
Jun 10, 2010 6:54:51 PM org.apache.catalina.startup.HostConfig deployDescriptor
INFO: Deploying configuration descriptor manager.xml from /var/lib/tomcat/conf/Catalina/localhost
Jun 10, 2010 6:54:51 PM org.apache.catalina.startup.HostConfig deployDescriptor
INFO: Deploying configuration descriptor hudson.xml from /var/lib/tomcat/conf/Catalina/localhost
Jun 10, 2010 6:54:52 PM org.apache.catalina.startup.HostConfig deployDescriptor
INFO: Deploying configuration descriptor ROOT.xml from /var/lib/tomcat/conf/Catalina/localhost
Jun 10, 2010 6:54:52 PM org.apache.coyote.http11.Http11AprProtocol init
INFO: Initializing Coyote HTTP/1.1 on http-8080
Jun 10, 2010 6:54:52 PM org.apache.coyote.http11.Http11AprProtocol start
INFO: Starting Coyote HTTP/1.1 on http-8080
Jun 10, 2010 6:54:52 PM org.apache.coyote.ajp.AjpAprProtocol init
INFO: Initializing Coyote AJP/1.3 on ajp-8009
Jun 10, 2010 6:54:52 PM org.apache.coyote.ajp.AjpAprProtocol start
INFO: Starting Coyote AJP/1.3 on ajp-8009
Jun 10, 2010 6:54:52 PM org.apache.catalina.startup.Catalina start
INFO: Server startup in 1178 ms
Jun 10, 2010 6:55:00 PM org.apache.catalina.core.ApplicationContext log
INFO: HTMLManager: init: Associated with Deployer 'Catalina:type=Deployer,host=localhost'
Jun 10, 2010 6:55:00 PM org.apache.catalina.core.ApplicationContext log
INFO: HTMLManager: init: Global resources are available
Jun 10, 2010 6:55:00 PM org.apache.catalina.core.ApplicationContext log
INFO: HTMLManager: list: Listing contexts for virtual host 'localhost'
Jun 10, 2010 6:55:35 PM org.apache.coyote.http11.Http11AprProtocol pause
INFO: Pausing Coyote HTTP/1.1 on http-8080
Jun 10, 2010 6:55:35 PM org.apache.coyote.ajp.AjpAprProtocol pause
INFO: Pausing Coyote AJP/1.3 on ajp-8009
Jun 10, 2010 6:55:36 PM org.apache.catalina.core.StandardService stopInternal
INFO: Stopping service Catalina
Jun 10, 2010 6:55:36 PM org.apache.coyote.http11.Http11AprProtocol destroy
INFO: Stopping Coyote HTTP/1.1 on http-8080
Jun 10, 2010 6:55:36 PM org.apache.coyote.ajp.AjpAprProtocol destroy
INFO: Stopping Coyote AJP/1.3 on ajp-8009
localhost_access_log.2010-06-10.txt
127.0.0.1 - z [10/Jun/2010:18:55:00 +0100] "GET /manager/html/ HTTP/1.1" 200 11308
127.0.0.1 - - [10/Jun/2010:18:55:30 +0100] "GET /hudson HTTP/1.1" 302 -
127.0.0.1 - - [10/Jun/2010:18:55:30 +0100] "GET /hudson/ HTTP/1.1" 404 982

jieryn added a comment - 11/Jun/10 10:26 AM

I gave this a try on Tomcat-7.0-RC4 and it seems that something strange is going on with the main welcome page. I can access other pages in Hudson just fine, e.g. http://localhost:8080/hudson/manage but the root welcome page is not loading. I looked back through a few versions previous to 1.361 and none of them define a welcome page..

Also, this exact version, 1.361, works fine on Tomcat-6.0.18. Someone more familiar with Hudson's crazy bootstrapping is going to need to get involved.


mindless added a comment - 11/Jun/10 01:09 PM

There may be crazy parts of Hudson bootstrapping but I don't think this part is crazy. war/resources/WEB-INF/web.xml is pretty basic, with:

<servlet-mapping>
    <servlet-name>Stapler</servlet-name>
    <url-pattern>/</url-pattern>
  </servlet-mapping>

which should map everything to stapler for handling..


jieryn added a comment - 11/Jun/10 01:36 PM

Change this to

<servlet-mapping>
  <servlet-name>Stapler</servlet-name>
  <url-pattern>/*</url-pattern>
</servlet-mapping>

and it will work. I tested this and things appear to be working sanely now.


kohsuke added a comment - 11/Jun/10 01:46 PM

According to the servlet 2.5 spec section SRV.11.2,

  • A string containing only the '/' character indicates the "default" servlet of the application. In this case the servlet path is the request URI minus the context path and the path info is null.

So I claim that '/' is valid according to the servlet spec.


pzz added a comment - 11/Jun/10 01:56 PM

I have also reported it in tomcat bugzilla, see https://issues.apache.org/bugzilla/show_bug.cgi?id=49422


mindless added a comment - 11/Jun/10 02:00 PM

Per Kohsuke's comment above, "/" is valid per the servlet spec. We hope that Tomcat 7 will fix this prior to their next beta/final release (or we'll surely hear about this again!)


jieryn added a comment - 29/Jun/10 07:10 AM

Tomcat closed that issue as INVALID, they believe they are strictly adhering to the specification. Specification debates aside, is there some reason we can not just migrate to /* ? I tested this in Tomcat 7 and it works properly, but did not check embedded Winstone, or any other popular containers.

Tomcat 7 has some unusual naming conventions for its releases, but they just announced a 7.0-BETA. The release manager stated that he'd like to get the first stable release out in 4 weeks or less. http://www.tomcatexpert.com/blog/2010/06/29/apache-tomcat-7-has-been-released

So I expect we're going to see a lot more Hudson users deploying to Tomcat 7 soonish... It would be nice if we didn't have to answer mailing list messages or spurious bug reports about Hudson not working.


jieryn added a comment - 29/Jun/10 07:10 AM

See comment above.


kohsuke added a comment - 21/Jul/10 07:32 PM

I still don't agree that this is our fault, and I've updated https://issues.apache.org/bugzilla/show_bug.cgi?id=49422 accordingly.

But in the mean time, I guess it'd wise to work around this issue in Hudson.

I'll try the "empty welcome file list" approach that Mark suggested to see if that does the trick, as that seems least likely to cause regressions with other containers.


kohsuke added a comment - 21/Jul/10 07:48 PM

Empty welcome file list didn't work. Apparently the default web.xml in Tomcat also must have no welcome file list.


scm_issue_link added a comment - 21/Jul/10 08:14 PM

Code changed in hudson
User: : kohsuke
Path:
trunk/hudson/main/war/resources/WEB-INF/web.xml
trunk/www/changelog.html
http://hudson-ci.org/commit/33149
Log:
[FIXED HUDSON-6738] changed the servlet mapping to /* to make it work with Tomcat. My fingers are crossed that this doesn't cause regressions elsewhere.


rdesgroppes added a comment - 27/Aug/10 06:34 AM - edited

Using a proxied Winstone with Security Realm set to "Delegate to servlet container", I had to restore:

<servlet-mapping>
  <servlet-name>Stapler</servlet-name>
  <url-pattern>/</url-pattern>
</servlet-mapping>

... otherwise (with /*) hudson.security.BasicAuthenticationFilter::doFilter() works with an empty path, with the effect that my browser faces a infinite redirect loop.

This is no surprise, according to javax.servlet.http.HttpServletRequest::getServletPath():

    Returns:
        a String containing the name or path of the servlet being called, as specified in the request URL, decoded, or an empty string if the servlet used to process the request is matched using the "/*" pattern.

Interestingly, I'm referring to Tomcat 7 hosted Servlet documentation: http://tomcat.apache.org/tomcat-7.0-doc/servletapi/javax/servlet/http/HttpServletRequest.html#getServletPath%28%29

In my opinion, the workaround should be reverted since it comes from a misinterpretation.