0000350: Resin doesn't start with httpd.sh -start switch

(0000398)
user347
08-12-05 00:00

Same thing happens with Resin 3.0.13 and 2.1.16 on Fedora Core 4. Works fine without the "start" switch. and with it just screws up.

(0000399)
ferg
08-12-05 00:00

Changing priority since this is with a new operating system and has been working without modification with several other unix operating system for years.

(0000400)
user352
08-12-05 00:00

We have this problem with Fedora Core 3, and a 'nasty' workaround - I hope this helps:
-An update containing IPv6 was installed on the FC3 server.
-I tried starting resin (both resin-3.0.14 and resin-3.0.12) and got the same result (nothing in the resin stderr or stdout logs, and a java <defunct> process) - and resin just tries restarting over and over.
-I modified wrapper.pl in resin/bin to comment out the following lines (~ line 712 - see bottom of my comment). This sent the error messages back to the console. I've lost the exact error messages, but it was along the lines that "Can't start java:... the child process needs to connect to the parent process, but socket=null". Sorry for not having the exact error msg :(
-The workaround was to remove Ipv6 from the server and it started fine with Ipv4.

resin/bin/wrapper.pl - line 712
if (($child = fork()) == 0) {
close(S);
# close(STDOUT);
# close(STDERR);
# open(STDOUT, ">>$jvm_log");
# open(STDERR, ">&STDOUT");

(0000401)
user347
08-12-05 00:00

Commenting out those lines seemed to have resolved the issue. Resin starts out fine with the start switch. Weird, eh. I am going to do some more testing see if i can pin point the problem to what is really causing the problem. I didn't hage the Can't start java error message though.

I'll comment if I find anything.

(0000402)
user372
08-12-05 00:00

I believe I found the cause...

if you run lsmod, you'll see the ipv6 module loaded. It seems to be a pretty low level module with (in my case) 36 dependencies. I renamed the module:

# cd /lib/modules/2.6.12-1.1398_FC4/kernel/net/
# mv mv ipv6.ko ipv6.ko.unused

then rebooted

wrapper.pl worked properly after doing this.

I can't imagine there being an issue with not loading this module.. if anything, less things to be worried about.

Brian

(0000403)
user372
08-12-05 00:00

A better option is to do this '
echo "alias net-pf-10 off" >> /etc/modprobe.conf

this'll prevent ipv6 module from loading.

This is a better method as the first will be overwritten with th enext kernel upgrade.

I tested this on a freshly installed FC4 box.

Brian

(0000404)
user347
08-12-05 00:00

Thanx Brian. I'll try this out.

(0000405)
ferg
08-12-05 00:00

moving to medium since there is a workaround.

Issue History
Date Modified	Username	Field	Change
08-12-05 00:00	user347	New Issue
08-21-06 08:44	ferg	Assigned To	=> ferg
08-21-06 08:44	ferg	Status	acknowledged => closed
08-21-06 08:44	ferg	Resolution	open => fixed
08-21-06 08:44	ferg	Fixed in Version	=> 3.1.0

Notes
(0000398) user347 08-12-05 00:00	Same thing happens with Resin 3.0.13 and 2.1.16 on Fedora Core 4. Works fine without the "start" switch. and with it just screws up.

(0000399) ferg 08-12-05 00:00	Changing priority since this is with a new operating system and has been working without modification with several other unix operating system for years.

(0000400) user352 08-12-05 00:00	We have this problem with Fedora Core 3, and a 'nasty' workaround - I hope this helps: -An update containing IPv6 was installed on the FC3 server. -I tried starting resin (both resin-3.0.14 and resin-3.0.12) and got the same result (nothing in the resin stderr or stdout logs, and a java <defunct> process) - and resin just tries restarting over and over. -I modified wrapper.pl in resin/bin to comment out the following lines (~ line 712 - see bottom of my comment). This sent the error messages back to the console. I've lost the exact error messages, but it was along the lines that "Can't start java:... the child process needs to connect to the parent process, but socket=null". Sorry for not having the exact error msg :( -The workaround was to remove Ipv6 from the server and it started fine with Ipv4. resin/bin/wrapper.pl - line 712 if (($child = fork()) == 0) { close(S); # close(STDOUT); # close(STDERR); # open(STDOUT, ">>$jvm_log"); # open(STDERR, ">&STDOUT");

(0000401) user347 08-12-05 00:00	Commenting out those lines seemed to have resolved the issue. Resin starts out fine with the start switch. Weird, eh. I am going to do some more testing see if i can pin point the problem to what is really causing the problem. I didn't hage the Can't start java error message though. I'll comment if I find anything.

(0000402) user372 08-12-05 00:00	I believe I found the cause... if you run lsmod, you'll see the ipv6 module loaded. It seems to be a pretty low level module with (in my case) 36 dependencies. I renamed the module: # cd /lib/modules/2.6.12-1.1398_FC4/kernel/net/ # mv mv ipv6.ko ipv6.ko.unused then rebooted wrapper.pl worked properly after doing this. I can't imagine there being an issue with not loading this module.. if anything, less things to be worried about. Brian

(0000403) user372 08-12-05 00:00	A better option is to do this ' echo "alias net-pf-10 off" >> /etc/modprobe.conf this'll prevent ipv6 module from loading. This is a better method as the first will be overwritten with th enext kernel upgrade. I tested this on a freshly installed FC4 box. Brian

(0000404) user347 08-12-05 00:00	Thanx Brian. I'll try this out.

(0000405) ferg 08-12-05 00:00	moving to medium since there is a workaround.

Relationships