lsphp issue

[IrPr]

Hi there,

Today i've upgraded one of servers to 4.1RC5 to use AIO support which fixes IO bottleneck and CPU is more utilized after that which is greatly appreciated
All websites now works like a charm and it really boosts!

But after few hours all php processes has been stopped working and processing suddenly and didn't proceed with any php script for minutes
CPU is fully idle and IOWAIT is 0 but no php process is handled

I just found this row in error log:
[notice] EACCELERATOR(17699): PHP crashed on opline 14 of read() at /home/lool/public_html/includes/core/core.cache.php:65

Then switched to XCache instead of EAccelerator, but after few more hours the same problem with php processes occurred without any error

I didn't check with static files if they are proceeded properly or not, but reloading/restarting lsws service fixed the problem
I'm not sure if its whole lshttpd issue or just for lsphp5 processes

For now i'm really confused
Any suggestion is welcome

Regards

[mistwang]

Do you have anything abnormal in error.log?
Do you get 503 errors?
Any lsphp process running?
Please enable debug logging with "killall -USR2 litespeed", make sure "Log Level" is "DEBUG", "Debug Level" is "NONE" in advance.
Also, you can run "strace -tt -T -p <pid_of_one_litespeed_child_process>", the output will help debugging the issue.

[IrPr]

Originally Posted by mistwang

Do you have anything abnormal in error.log?
Do you get 503 errors?
Any lsphp process running?
Please enable debug logging with "killall -USR2 litespeed", make sure "Log Level" is "DEBUG", "Debug Level" is "NONE" in advance.
Also, you can run "strace -tt -T -p <pid_of_one_litespeed_child_process>", the output will help debugging the issue.

I didn't find any 503 or 408 errors in access logs and Log Level it was set to "ERROR" which no any abnormal row was it in error.log

I see a few lsphp process running when it hung

DEBUG Log Level has been enabled to see whats wrong exactly,

but in fact i can't risk on this server because of customers that will get timeouts in case of this issue occurrence and added a cronjob to restart the lsws frequency

Do you have any suggestion to use another script instead of restart? It may doesn't log the reason of this issue but it restarts
I mean an script which can detect this issue very immediately and forces restarts then

Thanks in advance

[NiteWave]

since it looks like to be an opcode related issue, can you disable opcode cache and see if any improvement.