Daniel's questions

I have an apache instance running where 150 clients use long polling on some PHP scripts to receive updates about new files that have been uploaded. This of course creates 150 Apache threads. Now it seems the web server is at its limit (its an 10 year old machine) and I don't want to carry this baggage over to a new server.

I am thinking about using a WebSocket server that would handle the use case more effecient, but wonder, if 150 clients accessing the Apache which forwards the requests to my WebSocket server would also produce so many apache threads or if there is a more efficient way to handle it.

I cannot replace the Apache, just maybe with nginx because it also has to provide a homepage. I cannot use another port for the WebSocket server because this would mean I have to discuss update firewall rules with about 100 customers. Is this solvable?

I just installed two new virtual machines with Windows 8 and Windows 8.1 on our VMWare ESXi 5.1 Server and noticed a base load of about 800 Mhz without the machine doing actual work.

Now I wonder how one can reduce this to a minimum (e.g. my Linux fileserver has a minimum load of 6-10 Mhz, and Windows 2003 one of 40 Mhz). Believing Microsoft designed this OS to also run on mobile devices, there must be switches that stop the system from doing useless stuff. Any ideas?

I have created a Java Application that uses SWT 3.6 and now I want to use the browser widget on a 64bit Debian 6.0 system while using a 32bit Java JRE and the 32 bit SWT/GTK libraries.

I didn't feel this is a real programming question because I believe the server configuration part is the hard thing. If someone should guide me which libraries I have to install/configure to make this running it would be great.

I already installed 32bit GTK

apt-get install ia32-libs ia32-libs-gtk

and firefox

wget 'http://download.mozilla.org/?lang=de&product=firefox-18.0.1&os=linux'
mv index.html\?lang\=de\&product\=firefox-18.0.1\&os\=linux firefox-18.0.1.tar.bz2
tar xfvj firefox-18.0.1.tar.bz2
mv firefox /usr/local/
ln -s /usr/local/firefox/firefox /usr/local/bin/firefox

but still no luck.

Are there any open source firewalls for linux that

• can be installed on a server and bring their own linux OS?
• have a graphical user interface?

Would be great if the firewall

• would be based on Debian linux
• is still supported by an active community

The ones I found are all commercial and not suitable for usage in small enterprises.

PS: I know I could just fiddle with IPTables, but I wondered if there is a product that does not make me to do that.

Is it possible to make the Windows 2003 Server Dns Server lookup domains in the hosts file also? This would be nice to implement a corporate block list for all those ad sites, and for having a central place to configure them.

And the hosts file can be regulary updated from [http://winhelp2002.mvps.org/hosts.txt winhelp2002], so I don't have to enter 1000 sites manually.

I want to compile PostgreSQL and need to set a default for unix_socket_location. It should be /var/run/postgresql instead of /tmp like in the default build configuration.

In fact I just want to set a default for the unix_socket_location parameter, but I don't get where to set it. It is no ./configure option, and I don't know where to look anymore, althought the docs say it can be changed during build time.

Where do I set a default for unix_socket_location?

I have a subversion server on Windows (CollabNet download 1.6.16 on Windows Server 2003), and it regulary starts using up 100% CPU for hours. Sadly, there is no logfile for svnserve.exe, nor a possibility to see what he is working on (AFAIK).

How do i found out

• what he is doing?
• what he was doing?

Regards, Daniel

I want to measure some numbers related to disk io in a given timespan, programmatically. This means I don't want to spawn a process like iostat to get my numbers, by prefer to read something from proc or the like.

I need to know:

• The number of blocks written to disk
• The number of blocks read from disk
• The number of blocks read from disk but found in the disk cache
• The number of blocks about to be written to the disk, but not written yet, to know how much succeeding IO I might expect.

If I only get this information by disk or partition... this is even better, since I don't need to know how much information I read from proc.

If I additionally know how much time the process has waited for the IO to be done... even better.

Is this possible with some IO calls, or by reading and parsing the /proc files?

I have an Exchange Server 2010 running in a virtual machine on our VMWare Esx Server, and we do backups of the virtual machines regulary.

I heard somewhere that the Exchange Server should be stopped during the backup, because else I would risk a data loss. Is this true? Wouldn't this also imply that Exchange Server is not tolerant to a server power loss?

EDIT: I use VMWare ESX 3.5, and therefore do not have the vStorage API.

I have an exchange server 2010 running, and want to programmatically add contacts into the server. What is the way to achive this?

Best to be done in Java, but VB6 would be okay, too.

is there a direct download link to the JDKs and JREs? Maybe even a stable link to the latest JDK? I need to download Java at the command line and don't want to click through the Oracle homepage.

I have an VMWare ESX Server v3.5 with a few virtualized Debian Lenny VMs (kernel 2.6.22 with vmi) running Apache Tomcat 5.5. I enabled paravirtualization, and Disk IO increased from about 240MB/s to 380MB/s, making me a happy admin.

The problem now is that my apache tomcat becomes deadlocked during startup, running with 200% CPU (I have 2 CPUS assigned to the VM), and don't know how to get both: A stable system and a fast system.

I somewhere heared that paravirtualization is legacy anyway and won't be available on newer ESX servers. Is there a replacement for this seemingly performance-improving option, or is it discontinued becauses it is just unstable? What is the state of paravirtualization? Should I ignore it completely?

Thanks for all answers in advance.

I notices in the Network overview of Windows 7 that my switches are found on the path to the gateway, which baffled me completely. The last time I checked they just took the ethernet packages and transported them to the "right" location.

How does this switch detection work? Do switches alter the packages? Do they react on ICMP messages?