100% CPU Usage

Discussion in 'General' started by 1mb, Mar 19, 2010.

  1. 1mb

    1mb New Member

    Recently my webserver hanged few times in a week. After login into server to troubleshoot, found out the two lshttpd processes was 100% CPU utilised (refer attachment). Checked the access.log and error.log but seems like nothing unusual. After i excecute ./lswctrl stop and start again, it's seems like back to normal. Any advise?

    I'm using Litespeed Web Server Enterprise v4.0.13 on CentOS, running HTTP and HTTPS with Tomcat and PHP and equipped with 2Mbps dedicated international bandwidth.

    Attached Files:

  2. NiteWave

    NiteWave Administrator

    when lshttpd reach 100%, can you capture a few seconds what it's doing?

    #strace -p <pid-of-lshttpd>

    also, output of "pstree -p", lshttpd subtree may give some clue.
    Last edited: Mar 19, 2010
  3. 1mb

    1mb New Member

    Will check if it happens again. Thanks.
  4. 1mb

    1mb New Member

    It's happens again right now.. one of lshttpd is now 100%, tried to run "strace -p <pid>" but it seems like no output (refer Clipboard02.jpg) and please refer tree.txt for the "pstree -p" output.


    I also found out that I can't access my Litespeed Admin xxx.xxx.xxx.xxx:7080

    Millions thanks.

    Attached Files:

    Last edited: Mar 19, 2010
  5. NiteWave

    NiteWave Administrator

  6. 1mb

    1mb New Member

    4.1RC2 installed. Will monitor again.. thanks.
  7. mistwang

    mistwang LiteSpeed Staff

    to trouble shoot the 100% cpu issue, may need the debug build of lshttpd, and attach to the process with GDB.
    Please contact bug@litespeedtech... for further assistance if need.
  8. 1mb

    1mb New Member

    After upgraded to 4.1RC2. The 100% cpu usage issue stil happend quite frequent, after that the webserver not responding.

    Below is part of my litespeed setting:
    Max Connections : 350
    Connection Timeout (secs) : 100.
    Max Keep-Alive Requests : 150.
    Smart Keep-Alive : Yes
    Keep-Alive Timeout (secs) : 10

    The attached file was captured during the lshttpd 100% cpu usage.
    #strace -p <pid-of-lshttpd>

    Please advice.

    Attached Files:

  9. mistwang

    mistwang LiteSpeed Staff

    strace does not tell us what code causes the infinity looping, can you please download a debug build of LSWS then use GDB to attach to it, then use command "bt" to get callback stack trace. Please contact bug@litespeedtech... for the debug build of LSWS.
  10. 1mb

    1mb New Member

    Hi, I'd tried to send email to bug@litespeedtech.com but the email was bounced back with the following message "550 This message was classified as SPAM with score 6.0 (state 18)"

    Anyway, I've encounter one lshttpd 100% CPU usage at 6pm Malaysia time today and follow your instructions (please refer to my attachments).

    Usually after one lshttpd process is 100% CPU usage, I'm no longer able to access the Litespeed's admin web control panel and after a while (few minutes? few hours?) the 2nd lshttpd process will also reached 100% CPU usage and caused the web server hanged.

    From the gdb command, found this strange message "/home/gwang/release/httpd/httpd/extensions/jk/jconn.cpp: No such file or directory"

    Please advice, thanks.

    Attached Files:

  11. mistwang

    mistwang LiteSpeed Staff

    Please download and update to 4.0.14 release.
    You can get it by changing version number in the download link.
  12. 1mb

    1mb New Member

    But, currently I'm using 4.0.14 debug version.
  13. mistwang

    mistwang LiteSpeed Staff

    please download again.

Share This Page