Hi,

Since two weeks we have a problem with one of our websites, in „rush hours” (Analitycs shows ~170-200 people in real-time) it consumes all available memory (16GB).  Normal state is between 2-3GB allocated. Memory growth isn’t constant, sometimes is very rapid, like 4-16 GB/2 min, another time growth is gradual. This behaviour looks the same on both of our servers (Server1 and Server2).

Server configuration:

Traffic between public internet and the actual application servers is being handled by haproxy, currently all traffic is directed to Server1, if Server1 stops responding traffic goes to Server2 (active –backup configuration). Website database (MSSQL 2008 R2) is placed on Server3. On Media server we store all files from application virtual /media folder. That server is powered by Linux and no problems here.

Server1, Server2, Server3 are VMs placed on one physical machine (Linux Debian KVM,  latest version of libvirt from backports, machines are rock-stable, especially the DB Server). Media server is a physical storage machine.

Server1:

OS – Windows Server 2012 Standard

CPU – 8x2GHz

RAM -16GB

IIS8

Server2:

OS – Windows Server 2008 Web

CPU – 4x2GHz

RAM – 4GB

IIS7

Common things for both servers:

Site is based on Umbraco 4.7, .net 4.0. Media folder is connected as „network location”

physically placed on Media server (Linux Samba 3.x). In database we have about 25000 nodes.

We observed that website intensively use connection with Media server (up to 200Mbit/s).

We have changed the URLs so that media requests no longer pass through IIS.

Website was moved beetwen Windows Server 2008 and Windows Server 2012, yet the problem remains.

We thought that problem lies in the code, so we rolled back all the changes since last month (using our code repository), that didn’t solve the problem though.

We already have used programs such as DebugDiag and Ants Memory Profiler

How else can we check where the problem lies?