Kjetil Limkjær's questions

I've set up a DirectAccess server on Windows Server 2012 at my workplace. I'm using a Windows 8 Enterprise client to connect to it. It works fine over a mobile connection, but it fails when connecting from home. I've ruled out the firewall/router as the culprit as the issues persist when connecting the laptop directly to the cable modem.

I'm not sure where to begin to debug this, does anyone have any pointers? Both Teredo and IPHTTPS interfaces are up (although as the server is behind a NAT and we only have 1 public IP I understand that IPHTTPS is the only protocol that will be used).

The IPHTTPS tunnel also seems to be connected:

netsh interface httpstunnel show interfaces

Interface IPHTTPSInterface (Group Policy)  Parameters
------------------------------------------------------------
Role                       : client
URL                        : https://redacted:443/IPHTTPS
Last Error Code            : 0x0
Interface Status           : IPHTTPS interface active

however the DirectAccess link can't be activated - get-daconnectionstatus cycles between

Status    : Error
Substatus : CouldNotContactDirectAccessServer

and

Status    : Error
Substatus : RemoteNetworkAuthenticationFailure

Any suggestions on how to attack this are appreciated!

I'm browsing for a few low-end servers for a startup and I'm a bit confused about the different memory types. The advantage of ECC is clear - single-bit error correction. When it comes to registered memory it seems more vague, especially in systems that support both registered and unbuffered memory.

A Google search mostly finds copies of the Wikipedia article, which states that registered memory chips "...place less electrical load on the memory controller and allow single systems to remain stable with more memory modules than they would have otherwise". However I can't find any quantification of this.

What I'm wondering about is:

• Is registered memory an improvement over unbuffered when it comes to soft error rate, or is it purely about the maximum number of modules supported?
• If yes, at what point (amount of modules or GB of memory) do these improvements start to become noticeable?

For a specific example, the HP ProLiant DL 120 G6 server manual states that maximum supported memory configuration is 16 GB unbuffered (4x4GB) or 12 GB registered (6x2GB). In this case I'd rather have the extra 4GB of memory if the reliability difference is negligible.

I'm configuring a Hyper-V server (2008 R2) with two NICs. I'm trying to follow the MS best practice advice of leaving one NIC for the host OS. NIC #1 is used for the OS and has no associated virtual network. NIC #2 has a virtual network with "Allow management operating system to share this network adapter" unchecked. Both NICs are attached to the same physical switch.

From other PCs in the network, I can reach both the host and the guests without issue. However, between the host and the guests there is no connectivity.

Where should I begin to troubleshoot this?

We've been experiencing some problems with our Subversion server after upgrading to Ubuntu 10.04. When trying to access a repository, regardless of client (I've tried git-svn and svn on Windows as well as svn on Ubuntu 10.04, from different computers and network locations), I get a 400 bad request. Here's the output from svn:

svn: Server sent unexpected return value (400 Bad Request) in response to
     OPTIONS request for 'https://svn.example.org/svn/programs'

Here are the relevant entries from the Apache logs (I'm running Apache 2.2):

error.log

[Mon Jun 14 11:29:31 2010] [error] [client x.x.x.x] 
                           request failed: error reading the headers

ssl_access.log

x.x.x.x - - [14/Jun/2010:11:29:28 +0200] "OPTIONS /svn/programs HTTP/1.1" 
            401 2643 "-" "SVN/1.6.6 (r40053) neon/0.29.0"
x.x.x.x - - [14/Jun/2010:11:29:31 +0200] "ction-set/></D:options>OPTIONS 
            /svn/programs HTTP/1.1" 400 644 "-" "SVN/1.6.6 (r40053)
            neon/0.29.0"

If anyone has run into similar problems or could give me a pointer to track down the cause of this I'd be very grateful - I'd really like to avoid having to downgrade the box again.

Edit 2010-09-30:

There's a proposed fix in lucid-proposed now, apache2 2.2.14-5ubuntu8.3. See the bug report link in the accepted answer for details.

We've replaced our old file server with a new one running Windows Server 2008. The UPS we used for the old server is a GE NetPro 1500VA unit. It came with JUMP DataShield monitoring software, which seems to not have been updated since 2005. The most recent update I can find (2.2.0) does not support Windows 2008.

Does anyone know of alternative monitoring software that supports Windows 2008 and is compatible with the GE UPS? Although long power outages are rare here I'd still feel safer if I knew the server would shut down properly when the UPS battery runs low.

I'm having problems getting attachments to work in a new Trac installation. I'm running Trac 0.11.1 on Ubuntu 9.04. I'm relatively new to Trac, so hopefully I'm just missing something obvious.

After uploading an attachment, I'm sent to an empty list of attachments. The ./attachments folder in the Trac project directory is also empty. In addition, attachment links don't show up correctly in Trac - they just display as text, no link.

I've checked, and the Apache user does have write permission to the ./attachments folder. Could there be some .ini setting I've overlooked?

I've turned on DEBUG logging in Trac, but nothing regarding attachment uploading appears in the logs.