Re: Problems with IIS6.0 and UNC shares. Any help appreciated.

From: Bernard (qbernard_at_hotmail.com.discuss)
Date: 02/24/05 

Date: Thu, 24 Feb 2005 17:37:11 +0800

Followed up in directaccess group. 

-- 
Regards,
Bernard Cheah
http://www.tryiis.com/
http://support.microsoft.com/
http://www.msmvps.com/bernard/
"David Olive" <david.olive@westnet.com.au> wrote in message 
news:ugo7oLhGFHA.2740@TK2MSFTNGP12.phx.gbl...
> We are a second tier ISP based in Australia who has recently started 
> pushing Windows based hosting as an additional revenue stream.  We 
> currently host 700 websites on a load balanced IIS6.0 based server with a 
> NAS backend.  We are currently experiencing problems with our IIS 6.0 / 
> UNC configuration. Here is the background of our network layout.  Our NAS 
> is a Intel Dual Xeon server running Windows 2003 Standard edition and our 
> Web servers are both Pentium 3 1GHz machines running Windows 2003 Web 
> Edition.
>
>
>
> Requests come in on a load balancer that pushes each request to an IIS6.0 
> web server which then refers to a UNC share on our NAS for the data 
> backend. The load balancer polls each web servers default site every 15 
> seconds to ensure that the service is still up.  If it is unable to 
> retrieve a HTTP response header it will remove the server from the load 
> balancer.  Once this is done the load balancer continues to poll the 
> default site and once it is able to resolve it will be added back into the 
> cluster by the load balancer.
>
>
>
> We are finding that servers are being removed fairly often from the load 
> balanced cluster due to the load balancer being unable to retrieve the 
> default page.  Normally within a minute or so it is able to retrieve the 
> page and the server is then added back to the cluster.  Obviously this is 
> a problem due to session state being lost for our asp.net hosted 
> customers. We have done some testing of this behaviour and have noticed 
> that when a site is removed from the cluster we are unable to browse to 
> our NAS using UNC paths from both a management workstation or a web 
> server.  Once the cluster is added back the NAS responds almost 
> immediately.
>
>
>
> Based on this I assumed the problem may have something to do with SMB 
> queue requests but am having problems finding the exact settings to change 
> to fix or monitor the behaviour.  We have added some more powerful servers 
> into the equation but while this stops the problem occurring so regularly 
> it still seems to occur.  If the data is stored locally we have no 
> problems with removal of cluster members but this is not a solution to our 
> problem.  At the moment we have setup an IIS6.0 server running the 
> majority of our requests from local storage on our NAS, and also have our 
> other two servers configured with decreased weighting in our load 
> balancer.  I assume that the problem is no longer manifesting itself due 
> to the increased hardware and looking at the performance logs the majority 
> of the requests are being sent to our NAS which is serving from local 
> storage.  If the other two servers are given increased weighting then they 
> will start to drop from our load balancer again.
>
>
>
> If requested I am able to run an IISTrace and dump the web servers when 
> they are removed form the load balancer but I am not convinced the problem 
> lies with IIS6.0 as the servers are almost idle and IISCrash / Hang Agent 
> shows no errors.
>
>
>
> Sorry for the novel but I thought the more information the better.  Any 
> help with the above would be much appreciated, I'm ready to open a support 
> case but thought I would ask some of the newsgroups first.
>
>

Relevant Pages

  • Re: Clustering concepts...
    ... you can cluster to a shared resource on a NAS. ... 880 Filer and have been running SQL clusters quite nicely using Snapdrive ... > Replication is not a good technique for creating a warm standby server. ...
    (microsoft.public.sqlserver.clustering)
  • Problems with IIS6.0 and UNC shares. Any help appreciated.
    ... 700 websites on a load balanced IIS6.0 based server with a NAS backend. ... Our NAS is a Intel Dual Xeon ... The load balancer polls each web servers default site every 15 seconds to ... able to resolve it will be added back into the cluster by the load balancer. ...
    (microsoft.public.inetserver.iis)
  • Re: sql 2005 active/passive cluster options
    ... It just seems to me that using NAS with SQL Server in its current state is ... NAS may become robust. ... SQL Server MVP ... their product is an active/passive sql cluster. ...
    (microsoft.public.sqlserver.clustering)
  • Cheap Clustered Servers with 2003 possible?
    ... I want to build a 2003 server cluster, so that if I'm hosting a web ... Both machines would have fiber cards, while the NAS would have a fiber ...
    (microsoft.public.windows.server.clustering)
  • problem with rule failure on Sun B10n Load Balancer
    ... I am configuring a B10n Load Balancer from Sun, ... We are also running Oracle reports ... from a server external to the B1600 blade chassis. ...
    (comp.unix.solaris)