No web sites serving even though all services are active.

Discussion in 'General Questions' started by jamin, May 7, 2008.

  1. jamin

    jamin Kilo Poster

    Messages:
    72
    I have been experiencing a weird issue with a number of different Virtual servers and a number of different applications running on Virtual servers. It is happening on both Virtuozzo 3.5.1 and Virtuozzo 4.0. Both have latest service patches installed.

    The issue is intermittent and is not directly related to any action or resource issue on pysical server node. For some reason all web site in Plesk or a standalone IIS server including web application stops serving. VPS can ping with no issue, in fact you can RDP to server and all services are running. RAM on both VPS and physical node are not even close to being overloaded. If you shut down a single Virtual Server on node all sites come back.

    Has anyone seen same issue??? I surely cannot be the only one with this issue.

    HELP!!
     
  2. prakashkumaar

    prakashkumaar Bit Poster

    Messages:
    5
    We had similar issue and by investigating we could find that on the virtual container there are many application logs generated related to ASP net permissions for network service. Did you check the application event log on every virtual container that Event ID 10016 is exists? If so you can do the workaround mentioned in the Microsoft KB article
    http://support.microsoft.com/kb/931355

    After doing this we haven't faced the similar issue and not sure whether that was causing the problem really.
     
  3. jamin

    jamin Kilo Poster

    Messages:
    72
    I submitted a ticket to support team and got the following link to a KB article, i will have a look and try it out, I dont hold much luck however. It is interesting to note that they state that we shoudl nto run any more than 40 - 45 VE's on x86 server, totally opposite to their density model in their marketing material and licencing for VE's allow you to purchase 100 - 200 VE's. http://kb.odin.com/en/1675

    I will keep you posted on How I go...
     
  4. jamin

    jamin Kilo Poster

    Messages:
    72
    I check all servers as suggested by the KB and found zero errors, I also check individual VE's for any spyware as suggested and once again found nothing. Back to square one where sites and web applications will stop even though IIS is running. No fix at present.
     
  5. TarkanVeKurdu

    TarkanVeKurdu Kilo Poster

    Messages:
    54
    Hi ,

    I have the same issue on our 5 Virtuozzo servers with 3.51 and 4.0 . All patches are applied and systems are up2date. Every 5-10 days this problem occurs and web services stop responding. Rebooting the server made the system ok. Has anyone faced this problem and find a solution to this problem ??

    Regards
     
  6. jamin

    jamin Kilo Poster

    Messages:
    72
    I am still having issue. Support at Parallels suggested that it was page file however page file is set to a very large amount. I am at a loss to what it could be? Perhaps 32bit issue with Vituozzo and the way it handles memory and page file?
     
  7. mastermindpro

    mastermindpro Mega Poster

    Messages:
    117
    I was actually told that ~20 VE's on one 32bit node was pushing it. It sounds to me like their support department doesn't have a consistent story...yet again.

    I believe it all comes down to the nonpaged pool memory usage. We have management software that monitors that Perfmon counter on our 32bit nodes. We've never seen it max the 256MB limit, but we have seen the symptoms of it (event log messages indicating that a service failed to allocate memory).

    We're trying to move as much of our architecture to 64bit as possible, although Parallels should still hold an account for their "density statements" in marketing materials. Perhaps a class-action suit is in order...or maybe false advertising. We have a lot of attorney firms as clients. :^)
     
  8. jamin

    jamin Kilo Poster

    Messages:
    72
    Our nodes vary from 5 VE's (Containers) to 30 plus with same issue. The strange thing is that I did not see this issue in eairlier versions pre 3.5.1. I have ran Virtuozzo for Windows since RC1 days, I feel 64Bit may be next step. Will be interesting to see if others are having issue and have not reported it?
     
  9. bemerson

    bemerson Kilo Poster

    Messages:
    21
    This is no doubt a non-paged memory issue. Once you start hitting about 210-215 meg in the non-paged pool, services will start going flakey, especially IIS.

    Sometimes you can do a net stop/start on vzaop, or restart the service VE. This may buy you a day or two, but plan on restarting your hardware node a minimum of every 30 days to keep the non-paged pool in check.

    This by far is my biggest issue with Virtuozzo on Windows.
     
  10. mastermindpro

    mastermindpro Mega Poster

    Messages:
    117
    I agree with that...we've basically started doing just that on our 32bit nodes. Once we hit 215MB in the non-paged pool, paging goes through the roof and containers start getting flakey. It seems that we have to reboot our hardware node more often than every 30 days...more like once a week.
     
  11. ahphosting

    ahphosting Bit Poster

    Messages:
    3
    NonPaged memory and SmarterMail

    This is actually worse now than it used to be, and I can confirm from direct discussion with swsoft tech that this is NP pool issue. Now that being stated Parallels has a new partner, SmarterTools.com.

    Have a look at this thread from one of my post to the smartertools.com forums. http://www.smartertools.com/forums/t/11518.aspx

    No one with SmarterTools has responded to this post since I posted it 06-07-2008 10:40 AM
    I also have an OPEN support ticket with SmarterTools for this iisue and have simply been ignored.
    It is a shame that Parallels would market this software as Super High density and all I can get is 10 to 15 VEs on a 20 VE Lic.

    Greg Hedgepath
     
  12. jamin

    jamin Kilo Poster

    Messages:
    72
    Density is a major issue considering that the talk is HIGH density so you should be able to run more than 30 Containers on a single server running say Plesk, to find out that the large servers you spent a fortune on that have heaps of capacity are wasted because you can only fit half the number of Virtuals on server due to Pageing Issue.

    It is becomming an art not a science in depolying Containers, you need to mix the applications running on server and you can get some density but not what is promoted. I guess however you need to pick your alternative, right now, there really is none. Time will tell. Also do not use 64 bit version, you think you have issue now!!!
     
  13. mastermindpro

    mastermindpro Mega Poster

    Messages:
    117
    Don't use the 64bit version??? We've been using it for some time and it seems to have slightly fewer issues than the 32bit Virtuozzo. We're still in the process of re-building our containers to operate in the 64bit environment, but some of our containers have to remain 32bit. I think we can get down to 10 32bit containers, move our licenses around, and run like that for years to come.
     
  14. jamin

    jamin Kilo Poster

    Messages:
    72
    If you can let us know how you go with density on nodes with 64bit. Interesting to note that Parallels have scratched their code base and are starting from fresh with 64bit, one would wonder why? I also feel that 64bit is where the next step is however on our trials I have had less success with 64bit than 32bit. Any feedback on how you go would be super..
     
  15. mastermindpro

    mastermindpro Mega Poster

    Messages:
    117
    What about your 64bit environment hasn't worked? I have had one fewer problem in 64bit than I did with 32bit, namely the lack of such a low non-paged pool memory limit in 64bit. I would be very interested in hearing how much you know about their codebase, as you make some pretty significant claims about it.

    We have three hardware nodes running 64bit at the moment, and I haven't run into problems with any NPP memory limits on them.
     
  16. jamin

    jamin Kilo Poster

    Messages:
    72
    Ask the sales people about 64bit version and its all ok, but dig deeper and ask the techs and you will do a different story. I am not sure what status is on deveoplment of 64bit version all I can go by is what I have heard and have been told. As for the issues, they vary from service container constantly dropping ssh (very much like 32bit version) not a major issue unless you connect to node via PBA, which when service container corrupts itself you can connect and deploy subsciptions to node nor manage any containers from PBA. Most corruptions are caused by service patches. A few other issues such as no functioning terminal services from within a container, iscsi connectors dropping on container (I believe fixed now) and a number of smaller weird errors that I could not put my finger on such as slow perfomance on container yet same config and software running on second container runs quick, it just did not make any sense. I do plan to revist 64bit version at some stage, just need some time to test again.
     
  17. ahphosting

    ahphosting Bit Poster

    Messages:
    3
    SmarterMail has released a new version that finaly addresses the Non Paged Memory issue.
    I had one Hardware node that was running 14 VPs Lic for 20. 16 GB Ram and I still had 6 GB Free
    But NP Pool was at the limit. 9 of the 14 VPSs had Smartermail installed and used as much NP Pool
    as the total of some VPSs that did not have itinstalled. Once I updated all my installs I was able to see
    the overall NP Pool drop from 235588 k down to 1901043.

    I can now add at least three more VPSs to this node.
    I posted a follow up to the post on this page http://www.smartertools.com/forums/t/11518.aspx

    Greg
     

Share This Page