Howdy, Stranger!

It looks like you're new here. If you want to get involved, click one of these buttons!


Unmanaged node
New on LowEndTalk? Please Register and read our Community Rules.

All new Registrations are manually reviewed and approved, so a short delay after registration may occur before your account becomes active.

Unmanaged node

My box on an openvz node from bluevm reboot periodically.

it is the reason. someone on the same node is running out fd.

It seems bluevm do not manage/monitor the node.

Comments

  • What are you doing that requires 140,000 files open?

  • open fd table is shared between all openvz guests and the host.
    Maybe a buggy DoS script? only who manage the node (if they did) and the stresser know.

  • @Silvenga said:
    What are you doing that requires 140,000 files open?

    This. Sounds like you maybe need to switch to dedi..

    Thanked by 1Silvenga
  • cause said: open fd table is shared between all openvz guests and the host.

    You can't see the shared count in the OpenVZ guest. You have 140,000 open on your container.

    Stop attempting to find fault in BlueVM. Assume this is your fault until otherwise proven. Looks like you have a runaway script or something. Going from < 1K files to 140K files in a day means something is wrong.

    If you don't believe the OpenVZ documentation I checked the open file count on a small OpenVZ container with RamNode:

    cat /proc/sys/fs/file-nr
    1536    0       3276133
    

    This shows that I have 1.5K files open on my container. If I check a KVM container I see this:

    cat /proc/sys/fs/file-nr 
    1024    0       48763
    

    These two numbers are close enough (I'm using the OpenVZ for web hosting, the KVM for backups). As you can see, I only have access to the open files for my container.

    Thanked by 1Razza
  • @Silvenga said:
    These two numbers are close enough (I'm using the OpenVZ for web hosting, the KVM for backups). As you can see, I only have access to the open files for my container.

    [root@dalXXX ~]# cat /proc/sys/fs/file-nr
    1248    0   3276168
    [root@dalXXX ~]# vzctl exec XXXX cat /proc/sys/fs/file-nr
    1248    0   3276168
    

    dalXXX -> hostnode
    XXXX -> container ID

  • @Silvenga
    In openvz, atleast on 2.6.32 branch, /proc/sys/fs/file-nr shows usage of a single shared table in the kernel.

    I'm also monitoring numfile in /proc/user_beancounters I cannot believe my box held 14k fds.

  • serverian said: serverian

    I could be wrong, please enlighten me. Here is my logic:

    I have two OpenVZ containers on the same node: n4 and n1. Salt runs commands in parallel. lsof has a slight lag from file-nr, fd resources can be reused when resources are released. I hope these shortcomings are insignificant.

    salt "n[1,4]" cmd.run "lsof | wc -l ; sysctl fs.file-nr" 
    
    n4:
        1104
        fs.file-nr = 1488   0       19791963
    n1:
        1117
        fs.file-nr = 1488   0       19791963
    
    

    lsof doesn't list all the fd, but it does list all the open files. If each file uses one fd than lsof should always be less than fd.

    lsof only lists files open by my container. If file-nr shows all allocated fd for the whole system (including every container) then shouldn't this file-nr count be significantly higher than lsof count? Do I have 1K open files (sockets, files, devices, programs, pipes, etc.), while everyone else on my host has less than ~300 opened, combined and including the host?

Sign In or Register to comment.