Slew of WordPress 503s Constantly with LiteSpeed

#1
We have the latest build as of Thursday and we are still seeing constant issues with WordPress and 503 errors. When are these going to get resolved? We constantly have to place these clients on Apache servers due to the issues.
 

anything

Well-Known Member
#3
sounds familiar to a problem I had with apc

doh, i cant post links
in the litespeed php forum look for "Litespeed is using APC beta"
 
#4
No APC here. I am about to just scratch LiteSpeed on all of our servers if this persists.

Code:
2010-11-11 16:16:24.165 [INFO] [120.38.149.75:1450-0#APVH_sub.domain.com] connection to [/tmp/lshttpd/APVH_domain.com_Suphp.sock.371] on request #0, confirmed, 1, associated process: 16238, running: 0, error: Connection reset by peer!
2010-11-11 16:16:24.220 [INFO] [120.38.149.75:1450-0#APVH_sub.domain.com] connection to [/tmp/lshttpd/APVH_domain.com_Suphp.sock.371] on request #0, confirmed, 1, associated process: 16239, running: 1, error: Connection reset by peer!
2010-11-11 16:16:24.220 [NOTICE] [120.38.149.75:1450-0#APVH_sub.domain.com] Max retries has been reached, 503!
2010-11-11 16:16:24.220 [NOTICE] [120.38.149.75:1450-0#APVH_sub.domain.com] oops! 503 Service Unavailable

sounds familiar to a problem I had with apc

doh, i cant post links
in the litespeed php forum look for "Litespeed is using APC beta"
 

NiteWave

Administrator
#5
can you give more detail about the 503 error. what's the frequency. I'd like to check one of the site which has most 503 errors and see what's the root cause.
 
#6
This is an extremely common issue with Fantastico fresh WordPress installs. It is very frequent and I need a resolution. This issue does not exist whatsoever on our Apache install servers.
 

mistwang

LiteSpeed Staff
#7
The log message shows that the PHP process crashed when processing the request.
give our staff a temp root login to check the 503 errors, there must be subtle difference between LiteSpeed PHP and Apache PHP which cause the error.
either memory limit, opcode cache or something, if it can be reliably reproduced, we can figure it out.
 

webizen

Well-Known Member
#11
For 503 errors, error log (error_log for cpanel, error.log for native lsws) should show that multiple attempts has been made to process the request, if all of them ends up with "connection reset by peer", it is likely a bug on PHP side, maybe PHP crashed.

The best way to find out why PHP crash is to let it dump core files. LSAPI_ALLOW_CORE_DUMP environment must be set for lsphp5 external app.

Run "ulimit -a" from a php test script to make sure core dump limit is high enough.
Code:
<pre>
<?php
$ulimit = `ulimit -a`;
echo "ulimit:\n".$ulimit;
?>
</pre>
You can try crash a PHP process by manually sending "SIGBUS" signal while it is processing a test script takes long time to finish, something like calling sleep(60). make sure a core file created under the same directory where the script located. Otherwise, need to check the permission mask of that directory.

Once got the core file, can check it with GDB. Hope this is helpful. :)
 

markb1439

Well-Known Member
#13
Update: We're seeing the 503s with the default permalink setting. When we change to another setting (e.g., %category%/$postname% or virtually anything similar), the 503s go away.

Some users, however, may want the standard setting. How can this be made to work with LiteSpeed?

Thanks,

Mark
 

mistwang

LiteSpeed Staff
#14
Just follow the procedure to let PHP dump a core file when crash, then try to figure out what cause it to crash.
Sometime, it is the memory limit of lsphp5 external app, increase the limit see if it helps or not, check error_log and stderr.log first.
 
Top