|
I've been experiencing cache clearing as well. I have configured Litespeed to let PHP/LSAPI to self-manage its instances:
Instances: 1
Suexec: on
Max connections: 64
PHP_LSAPI_CHILDREN=64
LSAPI_CHILDREN=64
From my experience with PHP-FPM (where opcode caches work perfectly) the process tree in litespeed should work as:
- lshttpd
-- lsphp5 process manager
--- lsphp5 (as user example1 - child process)
--- lsphp5 (as user example1 - child process)
--- lsphp5 (as user example2 - child process)
--- lsphp5 (as user example2 - child process)
where the lsphp5 process manager spawns lsphp5 processes running as suexec user example1 and example2 depending on the requests served and each one of them can read/write the process manager's shared memory segment that caches both example1 and example2 user files.
What I get on Litespeed is no lsphp5 process manager and many example1 processed that each one of them run their own mmap cache segment which in turn constitutes any opcode cache useless.
We *need* an operational opcode cache mechanism on Litespeed! Please fix it!
|