hmallett's questions

I have a service which clients connect to using TCP. Currently there are 14 instances of this service, which are load-balanced by haproxy. Users on the internet connect to this service, and I have no control over the user's hardware which connects to this service.

The daemon which runs the service has been rewritten, and what I want to achieve is having a small number of users connect to the new daemon, leaving the rest of the users on the old version. I want to be able to control which clients connect to the new daemon (we know which IP ranges clients will be connecting from).

Once a client has connected to the new daemon, we don't want them to connect to the old daemon.

So I guess what we want to do is have haproxy choose a specific backend, for clients connecting to the front-end from a specific range.

Can this be done, and if so, how?

We are currently using haproxy 1.5.8 (packaged with our Linux distro), though we wouldn't rule out upgrading to a newer version if needed.

I have a Debian wheezy server which has many sources configured under /etc/apt

There's seems to be a lot of duplicate sources, and I would like to simplify this. In order to avoid breaking anything, and miss any package upgrades, I don't want to remove any sources which are the only available source of a package.

I know that I can see the source of packages for which there are upgrades available, by using

apt-get upgrade --just-print

How can I list the sources of all installed packages?

I have a pile of LTO 2 tapes, and both an LTO 2 drive (HP Ultrium 460e), and an autoloader with an LTO 3 drive in (Tandberg T24 autoloader, with a HP drive).

Performance of the LTO 2 tapes in the LTO 2 drive is adequate and consistent. HP L&TT tells me that the tapes can be read and written at 64 MB/s, which seems in line with the performance specifications of the drive. When I perform a backup (over the network) using Symantec Backup Exec, I get about 1700 MB/min backup and verify speeds, which is slower, but still adequate.

Performance of the LTO 2 tapes in the LTO 3 drive in the autoloader is a different story. HP L&TT tells me that the tapes can be read at 82 MB/s and written at 49 MB/s, which seems unusual at the write speed drop, but not the end of the world. When I perform a backup (over the network) using Symantec Backup Exec though, I get about 331 MB/min backup speed and 205 MB/min verify speeds, which is not only much slower, but also much slower for reads than for writes.

Notes: The comparison testing was done on the same server, SCSI card and SCSI cable, with the same backup data set and the same tape each time. The tape and drives are error-free (according to HP L&TT and Backup Exec). The SCSI card is a U160 card, which is not normally recommended for LTO 3, but we're not writing to LTO 3 tapes at LTO 3 speeds, and a U320 SCSI card is not available to me at the moment.

As I'm scratching my head to determine the reason for the performance drop, my first question is:

While LTO drives can write to the previous generation LTO tapes, does doing so normally incur a performance penalty?

I have an issue with an internally-deployed intranet site (so no links, sorry), and Internet Explorer 9.

The site is a Rails 3 app, using HTML 5, with nothing out of the ordinary - no javascript, just html and CSS. The HTML successfully validates (using the W3C validator).

The users on site all have either Internet Explorer 8 on Windows XP or Internet Explorer 9 on Windows 7. Users with IE8 are able to use the site as expected, as am I using Chrome.

Users who have IE9 though are able to view pages, but if they try and submit a form, IE loads a blank page, where the source code contains only html, head and body tags. They then find that they can't view any page on the intranet, and the IE back button doesn't work. If I use tcpdump on the intranet server, I can see that when the user clicks submit, no data is sent over the wire from the client, so I can only assume that the blank page (and its source code) is generated by IE.

I have ensured that IE detects the site as being in the "Intranet Zone", and that that zone has the default security setting of Medium-Low. I have tried with Compatibility Mode on (the default for intranet sites) and with it off.

To make matters worse, the behaviour is inconsistent - the form is successfully submitted about 5% of the time, and the rest of the time we have the above result.

I feel like I'm missing something fundamental - what am I missing?

I've got a Windows batch script which I'm trying to get Veritas Cluster Server to control using the process agent.

The batch script needs to run as a particular user, but when I try to online the resource I get the following error in the event log:

Source: AgentFramework

Event ID: 6532

Process:<<resource name>> Failed to launch the program <<path to batch script>>. Error = 1385

When I look up that error code here, it shows as:

Logon failure: the user has not been granted the requested logon type at this computer.

So the user account in question can't log on. That user account already runs some services, and has been granted the Logon as a service right. I've also added the Log on as a batch job right, but the problem persists.

So what rights do I need to assign to that account so that VCS can start a batch script as that user?

The operating system is Windows Server 2003, and VCS is version 5.0.

I have an ADSL connection which has a /29 subnet allocated to it, giving me 6 usable IP addresses. Currently this has a cheap Netgear ADSL router attached, which has a built in switch. There are 3 servers attached, each with a public IP address. Each of these servers is in our DMZ, and has a second network connection to the internal firewall, but I don't think that's important for this question.

Because the Netgear router's switch is so simple, each of these 3 servers can access the other servers, via the router's switch. What I require is for each of these servers to be isolated from the others, and have no access to them.

I intend to replace the router with something more suitable, such as a Cisco 1801, which also has a built-in switch, but supports VLANs on this switch. However, I'm not sure what the best method of achieving the goal is. I'm not sure if the firewall on that router applies to connections to its switch, or only routed connections. And I get the feeling that VLANs should be involved here, but I'm not sure how!

What is the best way for me to achieve the requirement of an ADSL connection with a /29 subnet, where the attached devices have no connectivity to each other?