High load

Discussion in 'Install/Configuration' started by ARPLink, Dec 20, 2012.

  1. ARPLink

    ARPLink Member


    In the last 24 hours, my server started to have an increased load time. Normally, it will stay around 3-4, but not it stays between 15-20.

    Here is a top and free output:

    Tasks: 271 total, 1 running, 270 sleeping, 0 stopped, 0 zombie
    Cpu(s): 14.7%us, 4.9%sy, 0.0%ni, 75.6%id, 3.9%wa, 0.0%hi, 0.8%si, 0.2%st
    Mem: 7340032k total, 6894408k used, 445624k free, 174848k buffers
    Swap: 4194296k total, 2736k used, 4191560k free, 4982420k cached

    root@s1 [/home/]# free -m
    total used free shared buffers cached
    Mem: 7168 6775 392 0 154 4931
    -/+ buffers/cache: 1690 5477
    Swap: 4095 2 4093

    My configuration is:
    Xeon Quad core
    7 Gb RAM

    Also, I would like to mention that EAProc WaitQ is always 0 for every site. Do you have any suggestions?

    Here are the console settings for LiteSpeed:
    Max Connections[B]=[/B]150
    Max SSL Connections[B]=[/B]100
    Connection Timeout (secs)[B]=[/B]300
    Max Keep-Alive Requests[B]=[/B]50
    Smart Keep-Alive[B]=[/B]Yes
    Keep-Alive Timeout (secs)[B]=[/B]2
    Send Buffer Size (bytes)[B]=[/B]0
    Receive Buffer Size (bytes)[B]=[/B]0
    I/O Event Dispatcher[B]=[/B]epoll (Linux 2.6 kernel)
    Max Cached Small File Size (bytes)[B]=[/B]8k
    Total Small File Cache Size (bytes)[B]=[/B]50M
    Max MMAP File Size (bytes)[B]=[/B]2M
    Total MMAP Cache Size (bytes)[B]=[/B]200M
    Use sendfile()[B]=[/B]Yes
    Use Asynchronized disk I/O (AIO)[B]=[/B]Keep File Mapped (64 bit OS)
    AIO Block Size[B]=[/B]8M
    No I-Node in ETag[B]=[/B]Not Set
    SSL Hardware Accelerator[B]=[/B](built-in) OpenSSL internal engine
    Max Request URL Length (bytes)=8K
    Max Request Header Size (bytes)=16380
    Max Request Body Size (bytes)=500M
    Max Dynamic Response Header Size (bytes)=8K
    Max Dynamic Response Body Size (bytes)=500M
    GZIP Compression	  
    Enable Compression=Yes
    Enable Dynamic Compression=Yes
    Compression Level (Dynamic Content)=4
    Compressible Types=text/*, application/x-javascript, application/xml,
    Auto Update Static=Yes
    Static GZIP Cache Directory=Not Set
    Compression Level (Static File)=4
    Max Static File Size (bytes)=1M
    Min Static File Size (bytes)=32K
    Per Client Throttling   
    Static Requests/second	=20
    Dynamic Requests/second=5
    Outbound Bandwidth (bytes/sec)=0
    Inbound Bandwidth (bytes/sec)=0
    Connection Soft Limit=10
    Connection Hard Limit=50
    Block Bad Request=Yes
    Grace Period (sec)=2
    Banned Period (sec)=300
    External Application Resource
    CGI Daemon Socket=Not Set
    Max CGI Instances=40
    Minimum UID=11
    Minimum GID=10
    Force GID=Not Set
    CPU Soft Limit (sec)=300
    CPU Hard Limit (sec)=600
    Memory Soft Limit (bytes)=100M
    Memory Hard Limit (bytes)=150M
    Process Soft Limit=10
    Process Hard Limit=15
    LiteSpeed API App Definition
    Notes	=Not Set
    Max Connections=50
    Initial Request Timeout (secs)=60
    Retry Timeout (secs)=0
    Persistent Connection=No
    Connection Keepalive Timeout=2
    Response Buffering=No
    Auto Start=Yes
    Back Log=100
    suEXEC User=Not Set
    suEXEC Group=Not Set
    Run On Start Up=No
    Max Idle Time=2
    Memory Soft Limit (bytes)=500M
    Memory Hard Limit (bytes)=600M
    Process Soft Limit=10
    Process Hard Limit=15
  2. NiteWave

    NiteWave Administrator

    is too low. the default is 2000.
    the reason to change the default value?

    check if any cloudliunx limit set too low ?

    please paste a real-time stats screenshot, it'll help to identify the bottleneck, where the problem is, etc.
  3. ARPLink

    ARPLink Member


    Here's an output of real-time status:


    I modified Max Conn accordind to your indication, installed CageFS(from CloudLinux) and the load dropped, but looking at the image, it seems to be large for only 37 requests
  4. NiteWave

    NiteWave Administrator

    yes, "Http Used" is 37 in this snapshot.
    but in case ddos time for example, the "Http Used" may increase to 1000+ in a few seconds. in this case, the web server may not work properly and cause the load high for a long time. Leaving much buffer as possible often is a good practice.
  5. webizen

    webizen Well-Known Member

    for quad core server, 4-5 load avg isn't a big deal.
  6. ARPLink

    ARPLink Member


    Any suggestions on buffer size in MB based on the existing RAM memory(it hat 7 Gb allocated).


    I'm aware that that's not a big load. As I said in the first post, when I asked, the load was at a minimum of 15. My issue is that it still looks to be a slightly big load considering the low number of requests that LiteSpeed has(especially considering that MySQL is on a dedicated server)
  7. NiteWave

    NiteWave Administrator

    is ok. 0 means default OS buffer size.

Share This Page