2.0RC3 Admin Interface problem?

#1
On my Solaris 10 b63 box I tried upgrading from 2.0RC2 to 2.0RC3 to fix the problem I've been having with lscgid not reaping children.

I shut-down the 2.0RC2 instance, did a mv /usr/local/lsws /usr/local/lsws.bak, and did a clean install of 2.0RC3 into /usr/local/lsws. Started it up, went to the admin port, logged in, went to 'Server Configuration' and I notice the page never finishes rendering (mainly the imges of the tabs at the top). If I try to click on anything a new window pops up, but nothing ever happens. Try closing my browser and opening a new one, and now I can't even get to the admin port.

Wiped RC3 and did a reinstall just in case, and it does the same thing. Wiped the RC3 install and mv'd the RC2 one back and it works fine.

Did I do something wrong?

Thanks,
-Dustin
 

mistwang

LiteSpeed Staff
#2
Dustin,

Thank you for the bug report.
I tried the 2.0RC3 on our Solaris 8 test machine, and it works very well.

Could you please provide more information for analysis?
Please install 2.0RC3 into a different directory, and set <debugLevel> to "10" in lsws/conf/httpd_config.xml, set <logLevel> to "DEBUG" in lsws/admin/conf/admin_config.xml manually, if the admin interface does not work. Please send us the error log file - lsws/logs/error.log.
For the lscgid problem, just copy the one in RC3 over to lsws/bin, it should work fine.

Thank you very much for your help!

George
 
#3
mistwang said:
Dustin,

Thank you for the bug report.
I tried the 2.0RC3 on our Solaris 8 test machine, and it works very well.

Could you please provide more information for analysis?
Please install 2.0RC3 into a different directory, and set <debugLevel> to "10" in lsws/conf/httpd_config.xml, set <logLevel> to "DEBUG" in lsws/admin/conf/admin_config.xml manually, if the admin interface does not work. Please send us the error log file - lsws/logs/error.log.
For the lscgid problem, just copy the one in RC3 over to lsws/bin, it should work fine.

Thank you very much for your help!

George
I'll do this tonight when I get home and I'll let you know how it goes..

I have to say so far I LOVE LSWS. It rocks! :)

-Dustin
 
#4
mistwang said:
Dustin,

Thank you for the bug report.
I tried the 2.0RC3 on our Solaris 8 test machine, and it works very well.

Could you please provide more information for analysis?
Please install 2.0RC3 into a different directory, and set <debugLevel> to "10" in lsws/conf/httpd_config.xml, set <logLevel> to "DEBUG" in lsws/admin/conf/admin_config.xml manually, if the admin interface does not work. Please send us the error log file - lsws/logs/error.log.
For the lscgid problem, just copy the one in RC3 over to lsws/bin, it should work fine.

Thank you very much for your help!

George
Okay, I finally figured this out.

The problem only occurs when I enable HTTP piplining in Mozilla Firefox. I didn't notice this in RC2 since I had it set to use HTTPS. If I disable piplining in Firefox or enable HTTPS on the admin port, the problem goes away. I have the log files tar'd up if that will help.

So RC3 seems to be better than RC2 now, except for one little annoyance that I've seen. 'lswsctl stop' on RC2 would completely shut LSWS down. 'lswsctl stop' on RC3 seems to always leave 3 proceses left: 1 lscgid and 2 lshttp's. If I kill them, the lscgid dies immediately, and then the lshttpd's die in about 5-10 seconds. If I truss the lscgid process before I kill it, I can see it trying to kill() the min lshttpd process over and over (using SIG0?), but it doesn't work...

Thanks for the awesome web server!
-Dustin
 
#5
jailbird said:
mistwang said:
Dustin,

Thank you for the bug report.
I tried the 2.0RC3 on our Solaris 8 test machine, and it works very well.

Could you please provide more information for analysis?
Please install 2.0RC3 into a different directory, and set <debugLevel> to "10" in lsws/conf/httpd_config.xml, set <logLevel> to "DEBUG" in lsws/admin/conf/admin_config.xml manually, if the admin interface does not work. Please send us the error log file - lsws/logs/error.log.
For the lscgid problem, just copy the one in RC3 over to lsws/bin, it should work fine.

Thank you very much for your help!

George
Okay, I finally figured this out.

The problem only occurs when I enable HTTP piplining in Mozilla Firefox. I didn't notice this in RC2 since I had it set to use HTTPS. If I disable piplining in Firefox or enable HTTPS on the admin port, the problem goes away. I have the log files tar'd up if that will help.

So RC3 seems to be better than RC2 now, except for one little annoyance that I've seen. 'lswsctl stop' on RC2 would completely shut LSWS down. 'lswsctl stop' on RC3 seems to always leave 3 proceses left: 1 lscgid and 2 lshttp's. If I kill them, the lscgid dies immediately, and then the lshttpd's die in about 5-10 seconds. If I truss the lscgid process before I kill it, I can see it trying to kill() the min lshttpd process over and over (using SIG0?), but it doesn't work...

Thanks for the awesome web server!
-Dustin
Okay, after working on this for hours I've found this:

1) Disabling pipelining and increasing the soft connection limit w/ RC3 helps, but eventually it still gets to the point where the admin interface slows to a crawl and then eventually locks. I noticed when this happens it also affects the regular listener (eg, phpinfo.php).
2) w/ RC2, everything works fine w/ pipelining and w/o changing the soft connection limit.

I'm not quite sure what changed in RC2->RC3 that make such a dramatic difference. I can say that so far RC2 + lscgid from RC3 is fast and rock-solid stable.

If there's anyhing else I can do to help test this, please let me know.

-Dustin
 

mistwang

LiteSpeed Staff
#6
Dustin,

Thank you for the update, we will investigate along that route. Have you ever tried it with other browser? Have you tried RC3 on Linux?

Thanks,
George
 
Top