Weblogs for lykwydchykyn
#9
Posted by lykwydchykyn on Tue 3 Apr 2007 at 22:27
I have been searching off and on for a while now to find some good materials on what it takes to go totally Linux -- that is, a roadmap for a complete Linux infrastructure. There seems to be a dearth of "standard" material once you get beyond single-system administration.
I mean, take for example the recent news about the french MP's migrating to Ubuntu. With 1100+ workstations, I want to know:
1. What sort of single-sign on or Identity management solution are they going to use?
2. What sort of software deployment system?
3. Are they going to have any sort of desktop policies being deployed?
It seems like a lot of things are "roll your own" nowadays, there isn't really an integrated solution a-la NDS or AD.
Can anyone point me to some relevant material on Linux infrastructures? Case studies, how-tos, etc? I mean, I can find "how to setup LDAP", but not really anything relevant on how to use it, if you know what I mean.
I mean, take for example the recent news about the french MP's migrating to Ubuntu. With 1100+ workstations, I want to know:
1. What sort of single-sign on or Identity management solution are they going to use?
2. What sort of software deployment system?
3. Are they going to have any sort of desktop policies being deployed?
It seems like a lot of things are "roll your own" nowadays, there isn't really an integrated solution a-la NDS or AD.
Can anyone point me to some relevant material on Linux infrastructures? Case studies, how-tos, etc? I mean, I can find "how to setup LDAP", but not really anything relevant on how to use it, if you know what I mean.
#8
Posted by lykwydchykyn on Tue 31 Oct 2006 at 04:37
I have a sarge box at work that hosts our county library's website. It's not a great piece of hardware, but it only serves up a few static html pages; the interactive content is hosted on a solaris server.
It's worked ok for quite some time, but recently apache2 has been slowing down and locking up on us, such that we have to restart the process. The log files only indicate that we need to raise the "maxclients" value, which I did (from whatever the default is (20?) to 30, then later to 150 because the problems persisted).
Are sarge's default settings too conservative? Should I have to tweak these out for a relatively low-traffic site like this?
It's worked ok for quite some time, but recently apache2 has been slowing down and locking up on us, such that we have to restart the process. The log files only indicate that we need to raise the "maxclients" value, which I did (from whatever the default is (20?) to 30, then later to 150 because the problems persisted).
Are sarge's default settings too conservative? Should I have to tweak these out for a relatively low-traffic site like this?
#7
Posted by lykwydchykyn on Fri 15 Sep 2006 at 17:13
Our organization just bought an iSCSI device for backups and various other things; I need to get my sarge boxes on this thing, but I am just a moron when it comes to compiling stuff. Under my hands the "make" command is strictly for generating massive amounts of errors to the screen.
Does anyone know if there is a backported .deb for the iSCSI initiator?
Does anyone know if there is a backported .deb for the iSCSI initiator?
#6
Posted by lykwydchykyn on Wed 31 May 2006 at 17:48
I use ncpmount to connect to netware servers here at work. The trouble is, after an hour or so the connection just goes comatose and won't work. If I try to unmount it, I get a failure to disconnect.
So I wrote a little script that kills all the ncp processes, and I figured that had it fixed. WRONG!
If I do a netstat, I still have 1 connection listed for every ncp connection I've made (including the ones I've killed). So after a few days I can end up with around 20-30 connections. Doing a netstat with the -p flag (even as root), there is no process associated with the connection, and ps -A turns up no ncp processes.
If I try to do /etc/init.d/networking restart, I get an error that there are still shares mounted. If I do a force-reload, it reloads the networking, but the connections are still there.
The only way I've found to clear the connections so far is a reboot, and that just won't do.
Which brings me to my question: is there a way to manually kill a udp or tcp connection that persists even when the process is dead?
So I wrote a little script that kills all the ncp processes, and I figured that had it fixed. WRONG!
If I do a netstat, I still have 1 connection listed for every ncp connection I've made (including the ones I've killed). So after a few days I can end up with around 20-30 connections. Doing a netstat with the -p flag (even as root), there is no process associated with the connection, and ps -A turns up no ncp processes.
If I try to do /etc/init.d/networking restart, I get an error that there are still shares mounted. If I do a force-reload, it reloads the networking, but the connections are still there.
The only way I've found to clear the connections so far is a reboot, and that just won't do.
Which brings me to my question: is there a way to manually kill a udp or tcp connection that persists even when the process is dead?
#5
Posted by lykwydchykyn on Fri 19 May 2006 at 04:51
I've played around with putting two IP's on a NIC before (e.g. -- eth0 & eth0:1), but I never could get use out of it because whenever I do it it seems to cause a delay when accessing network services, either from the box or to the box.
Well, now I have had to implement it for a server for actual reasons, and I'm wondering what exactly is causing this delay. I've observed it on at least 3 debian boxes, so I know it's not just a hardware issue or anything.
The delay seems to happen just on first access; once the service is accessed, there is no slowness; just an initial delay of 10-30 seconds.
The two IP's are on the same block, if it makes a difference, though I've observed it in situations where the IP's were on different blocks and at least once where there were two NIC's on separate blocks on the machine.
Anyone know what's up? Do I need to put in static routes or something?
Well, now I have had to implement it for a server for actual reasons, and I'm wondering what exactly is causing this delay. I've observed it on at least 3 debian boxes, so I know it's not just a hardware issue or anything.
The delay seems to happen just on first access; once the service is accessed, there is no slowness; just an initial delay of 10-30 seconds.
The two IP's are on the same block, if it makes a difference, though I've observed it in situations where the IP's were on different blocks and at least once where there were two NIC's on separate blocks on the machine.
Anyone know what's up? Do I need to put in static routes or something?
#4
Posted by lykwydchykyn on Thu 18 May 2006 at 04:50
My workplace is devoutly Novell; in fact, I was the only one there who really was into Linux at all until Novell started pushing Suse. Now, everyone's on about Linux. Well, almost everyone; our network admin has yet to soften in his dislike of all things Linux, so we still run mostly Netware 6.5 with NCP shares and NDS and all that good old stuff.
Having become the "linux guy", who got pulled in whenever we needed a simple and free server put together, I've been putting debian out there because I felt it was the best for the job. But now, of course, the long-time Novell folks are stoked about Suse, and I just can't get into it, but the need has begun to arise to get my Debian servers talking to their Netware servers. Novell has released a "client for linux", which irks me a bit because it should be called "client for Suse" as it doesn't work with anything else. I tried installing it using alien (got Groupwise and ConsoleOne going that way), but it's apparently far to complex to do with alien, I think it even patches the kernel somewhat.
I have used ncpmount in the past, but it's kind of a pain to get working, seems to run horridly slow, and doesn't always stay connected for more than an hour or two if left idle.
All that to say, has anyone out there got extensive experience integrating Debian into a Novell network, and how have you done it?
Having become the "linux guy", who got pulled in whenever we needed a simple and free server put together, I've been putting debian out there because I felt it was the best for the job. But now, of course, the long-time Novell folks are stoked about Suse, and I just can't get into it, but the need has begun to arise to get my Debian servers talking to their Netware servers. Novell has released a "client for linux", which irks me a bit because it should be called "client for Suse" as it doesn't work with anything else. I tried installing it using alien (got Groupwise and ConsoleOne going that way), but it's apparently far to complex to do with alien, I think it even patches the kernel somewhat.
I have used ncpmount in the past, but it's kind of a pain to get working, seems to run horridly slow, and doesn't always stay connected for more than an hour or two if left idle.
All that to say, has anyone out there got extensive experience integrating Debian into a Novell network, and how have you done it?
#3
Posted by lykwydchykyn on Thu 27 Apr 2006 at 19:17
Always been curious about apt-get; what's the difference between "recommended" packages and "suggested" packages? The words themselves don't suggest any difference in particular...
#2
Posted by lykwydchykyn on Wed 12 Apr 2006 at 05:16
I run several debian servers at work and one at home, as I'm sure many here do; I also run debian or debian derivatives on my personal workstations, again like many of you do.
What I'm wondering is if anyone here has deployed debian or other linux workstations in a business, school, public, or similar environment, and what your experiences have been. How has it scaled? How have you dealt with requests for windows-only software? Is it less or more trouble for users, in the long run?
What I'm wondering is if anyone here has deployed debian or other linux workstations in a business, school, public, or similar environment, and what your experiences have been. How has it scaled? How have you dealt with requests for windows-only software? Is it less or more trouble for users, in the long run?
#1
Posted by lykwydchykyn on Mon 20 Mar 2006 at 22:01
I'm trying to develop a setup for a public computer based around debian or a debian variant suitable for a setting like a public library. Users will have access to web-browsing, office applications, and possibly some educational games or science-oriented programs.
I'm a KDE fan so I've tried doing this using kioskTool. Unfortunately, the results are unpredictable and it doesn't seem to work consistently (such as with the KDE menu-- I would edit and edit but the changes never stuck. Apparently it's a bug that is not addressed in the current version).
Anyone had any experience doing something like this? Can you point me to some resources?
And no, for the record I'm not interested in diskless terminals.
I'm a KDE fan so I've tried doing this using kioskTool. Unfortunately, the results are unpredictable and it doesn't seem to work consistently (such as with the KDE menu-- I would edit and edit but the changes never stuck. Apparently it's a bug that is not addressed in the current version).
Anyone had any experience doing something like this? Can you point me to some resources?
And no, for the record I'm not interested in diskless terminals.