Howdy, Stranger!

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


virtualizor KVM node and CentOS7
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.

virtualizor KVM node and CentOS7

Hello,

Just wondering if you are using virtualizor for managing your KVM node, are you running CentOS6 or CentOS 7 on node?

Any performance differences ?? Or any issues ??
How good is virtualizor for managing KVM??

Thanks

«1

Comments

  • Neither, Ubuntu.

    One of the awesome perks of Virtualizor is you can use Ubuntu as opposed to a rhel variation. Its been over a year since I last touched Virtualizor, but back then we migrated from SolusVM + CentOS 6 to Virtualizor + Ubuntu 14.

    Once we moved to Ubuntu we saw nice performance improvements all around, it was pretty awesome honestly.

    I never used it with CentOS 7, but IMHO Ubuntu is a better solution for virtualization than rhel. Virtualization tech is constantly updating, however rhel always seems to be busy with back patches as opposed to updating to a newer version (Yes, its a two edge sword, but still..).

    Thanked by 2Umair deadbeef
  • One reply only?

    Not many people using virtualizor??

  • SpartanHostSpartanHost Member, Host Rep
    edited December 2016

    We've used Virtualizor on both CentOS 6 and 7. I'd recommend CentOS 6, 7 has a bug with libvirt where it crashes every few days meaning VPS's are no longer controllable until 'service libvirtd restart' is issued on the node. Virtualizor has confirmed there is no fix for this bug from libvirt yet. Only downside to using CentOS 6 vs 7 is the loss of I/O control per VPS.

    Never tried Ubuntu with Virtualizor before so can't comment on that.

    Thanked by 1Umair
  • AnthonySmithAnthonySmith Member, Patron Provider
    edited December 2016

    SpartanHost said: 7 has a bug with libvirt where it crashes every few days meaning VPS's are no longer controllable until 'service libvirtd restart' is issued on the node. Virtualizor has confirmed there is no fix for this bug from libvirt yet

    What is the bug link/report number?

    I think someone may be yanking your chain a bit.

    CentOS 7 is going to give you better performance, virtualizor will have no impact on performance it is only a control panel.

    I would rather use CentOS 7 with a mainline kernel than Ubuntu for long term stable VPS hosting any day.

    Thanked by 1Umair
  • A year ago I was using Virtualizor on Ubuntu, but had problems getting IPv6 to work. So I moved to CentOS 7 and everything works fine. Newer kernels give better performance especially if you're using SSDs.

    Thanked by 1Umair
  • @jesin said:
    A year ago I was using Virtualizor on Ubuntu, but had problems getting IPv6 to work. So I moved to CentOS 7 and everything works fine. Newer kernels give better performance especially if you're using SSDs.

    You are not seeing issue that @SpartanHost mentioned above ??

  • SpartanHostSpartanHost Member, Host Rep

    @AnthonySmith said:

    SpartanHost said: 7 has a bug with libvirt where it crashes every few days meaning VPS's are no longer controllable until 'service libvirtd restart' is issued on the node. Virtualizor has confirmed there is no fix for this bug from libvirt yet

    What is the bug link/report number?

    I think someone may be yanking your chain a bit.

    CentOS 7 is going to give you better performance, virtualizor will have no impact on performance it is only a control panel.

    I would rather use CentOS 7 with a mainline kernel than Ubuntu for long term stable VPS hosting any day.

    http://i.spartanhost.net/J9_c4y_a.png is the ticket we had with Virtualizor. I can't find any CentOS 7 bug reports right now but when we started having the problem, I think I was potentially able to find one that closely described what we were seeing. Happened even on a newly installed CentOS 7 node a few weeks ago, it very randomly occurs, could be fine for a week or two then suddenly starts crashing daily for a few days then back to being fine for a few weeks. All CentOS 6.7/6.8 nodes have been rock solid, would love to use only CentOS 7 though if I found out a way to stop libvirtd becoming unresponsive.

    Virtualizor doesn't handle a single node with unresponsive libvirtd very well, anyone with VPS's on that node won't be able to load Virtualizor whatsoever while the slave is unresponsive, even if they have VPS's on another slave too.

    Thanked by 1Umair
  • I used CentOS 7 + Virtualizor for almost an year & never faced any serious issues including those mentioned by @SpartanHost except for few minor bugs in virtualizor interface which they keep fixing in every update. However my setup was non- commercial-hosting/personal so the opinion of others including hosting companies, who use it on much more nodes than I did, might be different.

  • @Umair said:

    @jesin said:
    A year ago I was using Virtualizor on Ubuntu, but had problems getting IPv6 to work. So I moved to CentOS 7 and everything works fine. Newer kernels give better performance especially if you're using SSDs.

    You are not seeing issue that @SpartanHost mentioned above ??

    Nope, never had any issues with libvirtd but the node is only 40% full so maybe @SpartanHost's node was under heavy load.

    SpartanHost said: http://i.spartanhost.net/J9_c4y_a.png is the ticket we had with Virtualizor.

    Their reply is vague, maybe @virtualizor can fill in.

  • SpartanHostSpartanHost Member, Host Rep

    @jesin said:

    Nope, never had any issues with libvirtd but the node is only 40% full so maybe @SpartanHost's node was under heavy load.

    No high load, if we had high load and it crashed then that would probably be my primary suspicion making it easy to resolve.

  • AnthonySmithAnthonySmith Member, Patron Provider

    Are you using xfs on your root partition by any chance?

  • SpartanHostSpartanHost Member, Host Rep

    @AnthonySmith said:
    Are you using xfs on your root partition by any chance?

    Nope, ext4.

  • @SpartanHost said:

    Virtualizor doesn't handle a single node with unresponsive libvirtd very well, anyone with VPS's on that node won't be able to load Virtualizor whatsoever while the slave is unresponsive, even if they have VPS's on another slave too.

    I had a ticket with them and I asked them about it. So far no reply. (They were replying quickly before)

    May be @virtualizor can give some insight here.

  • exception0x876exception0x876 Member, Host Rep, LIR

    @SpartanHost said:
    We've used Virtualizor on both CentOS 6 and 7. I'd recommend CentOS 6, 7 has a bug with libvirt where it crashes every few days meaning VPS's are no longer controllable until 'service libvirtd restart' is issued on the node. Virtualizor has confirmed there is no fix for this bug from libvirt yet. Only downside to using CentOS 6 vs 7 is the loss of I/O control per VPS.

    Not sure if it is the same bug, but I had a similar one on my CentOS 7 nodes with OpenNebula (which uses libvirt under the hood). Once in a few days libvirt became unresponsive showing 100% CPU usage.

    I've fixed it by setting different permissions on libvirt socket in libvirtd.conf. Since then all the nodes run smoothly with no problems at all.

  • SpartanHostSpartanHost Member, Host Rep

    @exception0x876 said:

    @SpartanHost said:
    We've used Virtualizor on both CentOS 6 and 7. I'd recommend CentOS 6, 7 has a bug with libvirt where it crashes every few days meaning VPS's are no longer controllable until 'service libvirtd restart' is issued on the node. Virtualizor has confirmed there is no fix for this bug from libvirt yet. Only downside to using CentOS 6 vs 7 is the loss of I/O control per VPS.

    Not sure if it is the same bug, but I had a similar one on my CentOS 7 nodes with OpenNebula (which uses libvirt under the hood). Once in a few days libvirt became unresponsive showing 100% CPU usage.

    I've fixed it by setting different permissions on libvirt socket in libvirtd.conf. Since then all the nodes run smoothly with no problems at all.

    What exactly did you set in the libvirtd.conf?

  • exception0x876exception0x876 Member, Host Rep, LIR

    @SpartanHost said:
    What exactly did you set in the libvirtd.conf?

    auth_unix_ro = "none"
    auth_unix_rw = "none"
    unix_sock_group = "oneadmin"
    unix_sock_ro_perms = "0770"
    unix_sock_rw_perms = "0770"
    

    "oneadmin" is the name of the group operating OpenNebula.

    Thanked by 1Umair
  • SpartanHostSpartanHost Member, Host Rep

    @exception0x876 said:

    @SpartanHost said:
    What exactly did you set in the libvirtd.conf?

    > auth_unix_ro = "none"
    > auth_unix_rw = "none"
    > unix_sock_group = "oneadmin"
    > unix_sock_ro_perms = "0770"
    > unix_sock_rw_perms = "0770"
    > 

    "oneadmin" is the name of the group operating OpenNebula.

    Thanks, will take a look later to see what ours are set to.

  • Hi,

    It seems to be related to the virt-df slowness.
    We have debugged this and will fix it.
    A patch should be launched in the next 1-2 days.

    Regards

  • AnthonySmithAnthonySmith Member, Patron Provider

    virtualizor said: virt-df slowness.

    I would suggest if using virt-df on a busy VG then it is an on demand thing i.e. a refresh button with a date/time stamp rather than an on page load, certainly for the admin interface anyway.

    Thanked by 1i83
  • @AnthonySmith said:
    i.e. a refresh button with a date/time stamp rather than an on page load, certainly for the admin interface anyway.

    I'd suggest this for a large portion of the querying for the admin interface.

  • @virtualizor said:
    Hi,

    It seems to be related to the virt-df slowness.
    We have debugged this and will fix it.
    A patch should be launched in the next 1-2 days.

    Regards

    Thanks for the update. At least you guys are working on improving / fixing things ... Unlike Solus ... :(

  • alex0alex0 Member
    edited December 2016

    @AnthonySmith said:

    virtualizor said: virt-df slowness.

    I would suggest if using virt-df on a busy VG then it is an on demand thing i.e. a refresh button with a date/time stamp rather than an on page load, certainly for the admin interface anyway.

    No, they dont load it in the admin interface. Its called during an hourly cron.

  • AnthonySmithAnthonySmith Member, Patron Provider

    alex0 said: No, we never load it in the admin interface. Its called during an hourly cron.

    You work for virtualizor?

    Why call all that info hourly at all, that is a lot of crap IO to generate for no good reason, that is probably the reason you are finding it 'slow', doing it hourly is even worse than on page load.

  • @AnthonySmith said:

    alex0 said: No, we never load it in the admin interface. Its called during an hourly cron.

    You work for virtualizor?

    Why call all that info hourly at all, that is a lot of crap IO to generate for no good reason, that is probably the reason you are finding it 'slow', doing it hourly is even worse than on page load.

    No I dont work for them. Sorry for english. Corrected it.
    I was the guy who reported it. They gave me reason that its need to give true disk utilization to users.

  • AnthonySmithAnthonySmith Member, Patron Provider

    alex0 said: They gave me reason that its need to give true disk utilization to users

    Well i understand that but still, no need to load it for admins unless it is requested, even for users the option to refresh/load on demmand would make more sense in terms of load on the server.

  • @AnthonySmith said:

    alex0 said: They gave me reason that its need to give true disk utilization to users

    Well i understand that but still, no need to load it for admins unless it is requested, even for users the option to refresh/load on demmand would make more sense in terms of load on the server.

    There is an option to disable it as well. It was asked by some webhosting companies, hence it was added.

  • AnthonySmithAnthonySmith Member, Patron Provider

    virtualizor said: There is an option to disable it as well. It was asked by some webhosting companies, hence it was added.

    Disable for users or disable for admins or both, and if disabled is there an option to load it on demand?

    Thanked by 1i83
  • @AnthonySmith said:
    Disable for users or disable for admins or both, and if disabled is there an option to load it on demand?

    OnDemand certainly make more sense for this. @virtualizor Considering giving options to admin on how to set this up.

  • UmairUmair Member

    Sorry bumping an old thread. Last time I went with CentOS 6.x as I was told
    "For best Stability, go with CentOS 6.8".

    Wondering if anythin has changed .... There have been few updates to @virtualizor panel. Are all issues fixed ??

    Last time the big NO for me was, if mian node reboots, VM wont come back online on their on with CentOS7.

  • jackbjackb Member, Host Rep

    We don't use virtualizor, but CentOS 7 as a KVM host has been perfectly stable for us.

Sign In or Register to comment.