| Re: High Load D states.. [message #34934 is a reply to message #34894] |
Thu, 12 February 2009 16:56   |
kevinm
Messages: 12
Registered: February 2009
|
Junior Member |
|
|
Hi there,
Thanks for the reply !
| Quote: |
Does removing setuid solve the problem in general?
|
Yes, removing the setuid from the process removes the 'wait' states introduced by this syscall / the delay in the syscall and solves the problem (however removes a large security layer)
| Quote: |
Does setuid32 function always have "77777777" as a parameter? Or this number always changes? Is 77777777 a valid number?
|
This number always changes with the user calling the script from the web server, and the number is valid (the uid is stat'd from the file system, however in the strace the stat is completed, and the number displayed BEFORE the delay, the delay occurs when doing the setuid)
The issue shows itself when the system is under a high request per second schenario (peak points of the day, however is fully reproducable , I will see if I can get some infor from salbinfo when this is occuring.
Is there any tunables that you are aware of for the uid_cache that I can look at ?
Again THANKS for the reply, this is causing me a lot of troubles.
Regards
Kev
|
|
|
|
OpenVZ Forum
Members
Search
Help
Register
Login
Home
