tomcat not working on the ISO version, ok on the VM version

[camerons]

I've tried running the rasdaman quickstart, latest version at: ​http://adhoc.osgeo.osuosl.org/livedvd/docs/en/index.html I'm using a modified version of osgeo-live5.0rc1 + latest rasdaman script.

Issues found ... From ​http://localhost:8080/earthlook/demos/time-series/value-retrieval.php :

• Click on ts_001 and get following error:

Error in executing query: ​http://kahlua.eecs.jacobs-university.de:8080/petascope/earthlook

• I get similar errors for other links ts_002, ...

From: ​http://localhost:8080/earthlook/demos/climate/usecase-slicing.php Click on one of the cubes, and nothing seems to happen. There is no user feedback.

[hamish]

notes: seems like tomcat6 is giving us problems right now.

we needed to set JAVA_HOME better, I put it in rc.local but that gets run not sourced so it wasn't exported back to the parent process. now it's in /etc/profile.d/ and hopefully it sticks.

if needed we can hardwire it in /etc/default/tomcat6

after fixing that, and setting sun's java to be the default with update-alternatives --config java, I still get a [fail] each time when I do /etc/init.d/tomcat6 start.

after changing the shebang in that script to #!/bin/sh -x the end of the output looks like:

+ sleep 5
+ start-stop-daemon --test --start --pidfile /var/run/tomcat6.pid --user tomcat6 --exec /usr/lib/jvm/java-6-sun/bin/java
+ [ -f /var/run/tomcat6.pid ]
+ rm -f /var/run/tomcat6.pid
+ log_end_msg 1
+ [ -z 1 ]
+ [ 91 ]
+ [ -x /usr/bin/tput ]
+ log_use_plymouth
+ [ n = y ]
+ plymouth --ping
+ printf \r
+ /usr/bin/tput hpa 91
                                                                                           + [ 1 -eq 0 ]
+ printf [
[+ /usr/bin/tput setaf 1
+ printf fail
fail+ /usr/bin/tput op
+ echo ]
]
+ return 1

??

also I notice at the end of /var/log/tomcat6/catalina.out

[...]
Aug 23, 2011 10:55:17 PM org.apache.catalina.startup.HostConfig deployWAR
INFO: Deploying web application archive wss.war
Aug 23, 2011 10:55:19 PM org.apache.catalina.loader.WebappClassLoader validateJarFile
INFO: validateJarFile(/var/lib/tomcat6/webapps/wss/WEB-INF/lib/servlet-api-2.4.jar) - jar not loaded. See Servlet Spec 2.3, section 9.7.2. Offending class: javax/servlet/Servlet.class
Aug 24, 2011 12:00:29 AM org.apache.catalina.startup.HostConfig checkResources
INFO: Reloading context [/host-manager]

the /var/lib/tomcat6/webapps/wss/WEB-INF/lib/servlet-api-2.4.jar file does exist in the filesystem, so I'm not sure if that is a useful clue or not..

??, Hamish

[hamish]

more broken tomcat comments:

looking at 5.0rc2, tomcat works on the VM and does not work on the ISO.

ie sudo /etc/init.d/tomcat6 start fails.

Alex has this theory, and I think he could be on to something:

what is different between the two? build_iso.sh.

in that we use fslint's findup program to find all duplicate files on the disc and hardlink them together. as there are known to be a few duplicate copies of tomcat around which diff't projects have self-bundled, it may be that something is getting tied together which shouldn't be.

afaict there is not a simple way to ask fslint to avoid a particular directory.

currently I'm hunting for the GNU find incantation to find all files within /usr/share/tomcat6/ and /var/share/tomcat6/ which are hardlinks.

help in the next few hours from someone/anyone with a clue about tomcat would be most appreciated.

thanks, Hamish

[hamish]

it turns out most of the tomcats on the ISO are non-functional.

this definitely affects 52n, and maybe deegree, geonetwork, mapfish, and rasdaman too.

It seems to be the same cause as the atlasstyler.sh bug: losing its +x permissions for unknown reasons.

Solution seems to be to do

chgrp tomcat6 *.sh

as it runs as the tomcat6 user, who was part of the 'chmod o-x' denied.

after chgrp to tomcat6 the /etc/init.d/tomcat6 start succeeds.

(and hope like hell it doesn't forget the group name when the build_iso runs)

there may be other problems, but this seems like it was a contributing factor.

Hamish

[hamish]

most seem functional in 5.0rc5 ISO except for deegree, which is colliding with the system's tomcat.

I've added modified start/stop deegree scripts in svn ./app-conf/deegree/ which turn off the system's tomcat when you start deegree, and restarts it when you stop deegree, which seems to solve the problem there.

Hamish

[darkblueb]

after running Start Rasdaman and Start Geomajas

  $ sudo /etc/init.d/tomcat6 status

returns a pid

appears to be fixed in RC6

[camerons]

Tried rerunning the rasdaman steps as described in the original bug report. Confirmed that rasdaman is working ok. Verified that a number of other java webapps are also working (tested 52N apps).

Used 5.0rc6 ISO in a VM with 768 Meg RAM.

Hence, closing this issue as fixed.