DF, MC very slow since VirtualFileSystem errors

  • IMSCP 1.3.7
    Debian 8


    When i execute the command df the o/p hangs for about 2 minutes,
    also MC is very slow and the Panel SysInfo page ...
    Could this be an issue w/ intenal imscp mount points and how can I solve the problem?


    CPU load looks good:
    Cpu(s): 2.0 us, 1.8 sy, 0.0 ni, 96.2 id, 0.0 wa, 0.0 hi, 0.0 s

  • it seems there is a loop running ...
    the output of df -a shows thousands of the same subdomain entrys

    Code
    1. df -a
    2. /dev/md2 483630616 147098292 311942244 33% /var/log/apache2/subdomain.domain.com
    3. /dev/md2 483630616 147098292 311942244 33% /var/log/apache2/subdomain.domain.com
    4. /dev/md2 483630616 147098292 311942244 33% /var/log/apache2/subdomain.domain.com
    5. /dev/md2 483630616 147098292 311942244 33% /var/log/apache2/subdomain.domain.com
    6. /dev/md2 483630616 147098292 311942244 33% /var/log/apache2/subdomain.domain.com
    7. /dev/md2 483630616 147098292 311942244 33% /var/log/apache2/subdomain.domain.com
    8. ...

    I was able to stop the loop by deleting the subdomain ..


    I will disable the edit button for subdomains until next release ;-)

    Edited once, last by fulltilt ().

  • @fulltilt


    It seem that there is multiple bind mounts for the same directory which should not occurs. Can I access to your server to check?


    BTW: You should really avoid to edit i-MSCP code. There is a service for mounts management.

    badge.php?id=1239063037&bid=2518&key=1747635596&format=png&z=547451206

  • Thank you Nuxwin!
    Yesterday I noticed that imscp_traffic was not started I have restarted the server and it looks good until now ... no more duplicate entries from df -a
    I have added a new customer too and everything seems to work.
    Maybe the reason for this is that imscp_traffic service was not started ...
    I guess it can stay that way until the next release.

    Edited 2 times, last by fulltilt ().

  • same problem occurs when changing PHP settings f.ex. switching error reporting on / off
    I can kill the mount loops w/
    umount -l /var/log/apache2/example.com
    but I'm not sure if the killed mountpoint needs to be re-created after that ... I guess it works without re-create it ...


    If there is a quick fix just point me to the right part which has to be changed ... I have some experience with Perl and PHP and I can handle w/ it.
    I do not want to disturb you at work.

  • I cannot reproduce the problem. Please show me your /etc/imscp/mounts/mounts.conf file content and also the result of the following command:


    Code
    1. # findmnt -o source,target,propagation


    through our conversation system if needed.

    badge.php?id=1239063037&bid=2518&key=1747635596&format=png&z=547451206

  • @fulltilt


    I did some tests. I created 4000 customer accounts, meaning 4000 bind mounts.


    Issue founds

    • The imscp_mountall command hangs (not really infact, process is just really slow) when there is duplicate bind mounts (I must investigate and fix)

    My test environment


    root@srv01:~# lsb_release -a
    No LSB modules are available.
    Distributor ID: Debian
    Description: Debian GNU/Linux 8.6 (jessie)
    Release: 8.6
    Codename: jessie
    VM with 4 CPU cores assigned and 2 Gio memory

    badge.php?id=1239063037&bid=2518&key=1747635596&format=png&z=547451206

  • Thanks Nuxwin,
    I have tested in a V-Box copy of this server and added a new customer ...
    Is this still a correct mount point even if it's shown duplicated from output df -a?

    Code
    1. /dev/sda1 24606076 11873940 11459152 51% /var/www/virtual/test.gr/logs/test.gr
    2. /dev/sda1 24606076 11873940 11459152 51% /var/www/virtual/test.gr/logs/test.gr
    3. /dev/sda1 24606076 11873940 11459152 51% /var/log/apache2/test.gr
  • @fulltilt


    No, that is not ok. Please, run the attached script on your vbox as root user (perl umount.pl). This will remove any bind mount from /var/www/virtual (recursively). Be aware that this can take several minutes (depending of number of duplicate bind mounts).


    Once done

    • Rerun the i-MSCP installer from the i-MSCP 1.3.7 archive.
    • Add a subdomain and edit it

    Once done


    Say me if there is still duplicate entries.

    Files

    • umount.pl

      (204 Byte, downloaded 5 times, last: )

    badge.php?id=1239063037&bid=2518&key=1747635596&format=png&z=547451206

  • OK, directly after installation / upgrade (incl. umount) all domains, subdomains and alias domains are duplicated:

    Code
    1. df -a/dev/sda1 24606076 11882880 11450212 51% /var/www/virtual/example.com/logs/alias.com/dev/sda1 24606076 11882880 11450212 51% /var/www/virtual/example.com/logs/alias.com

    At the second test after umount and reboot, everything looks good ...
    But if I add a new subdomain


    Code
    1. df -a/dev/sda1 24606076 11883720 11449372 51% /var/www/virtual/test.gr/logs/null.test.gr/dev/sda1 24606076 11883720 11449372 51% /var/www/virtual/test.gr/logs/null.test.gr/dev/sda1 24606076 11883720 11449372 51% /var/log/apache2/null.test.gr


    and if I change and safe PHP settings it becomes very bad

    hmm, possibly caused by the opcode-cache?


    Regards