Hello
I have 2 problems with udpcast currently:
1. I need to make the nodes reboot automatically efter udprcv has run. (Otherwise I'll need a ladder to go around and reboot 36 nodes :) ) I don't know how I do this I've tried unzipping and "un-cpio-ing" the initrd but averything seems to be embedded in busybox, so I don't know what to add. Is there a "recipe" for this somewhere.
2. We are using a Cisco switch here, which has an annoying habbit of closing ports once linux releases the device just prior to a dhcp discover request. The port won't get opened till at least 5-6 seconds later. Hence udpcasts automatic dhcp-attempt fails. Is there a way to increase the number of tries, or the timeout? The netwait= option is no good here, because the port gets closed just when the dhcprequest is initialised, so I need 5-6 secs from there. And this I guess can only be done with a higher number of tries or a higher timeout.
Lasse Riis
I don't know about the Cisco switch aspect, but we had a similar problem with buggy tg3 ethernet device drivers which sounds similar. Our solution was to use ipappend 1 at the head of the default file for the udpcast PXE boot. This forces it to reuse the initial DHCP address rather than request it again. The other thing that might be helpful is the command line option --rexmit-hello-interval for udp-sender. You might also want to experiment with starting the receivers or udp-sender first and see if one works better for you.
The reboot should be possible by making a udpreceiver.post shell script within your image. Busybox includes a reboot command, which may work for you.
--Donald Teed
On Tue, 28 Jun 2005, Lasse Riis wrote:
Hello
I have 2 problems with udpcast currently:
- I need to make the nodes reboot automatically efter udprcv has run.
(Otherwise I'll need a ladder to go around and reboot 36 nodes :) ) I don't know how I do this I've tried unzipping and "un-cpio-ing" the initrd but averything seems to be embedded in busybox, so I don't know what to add. Is there a "recipe" for this somewhere.
- We are using a Cisco switch here, which has an annoying habbit of
closing ports once linux releases the device just prior to a dhcp discover request. The port won't get opened till at least 5-6 seconds later. Hence udpcasts automatic dhcp-attempt fails. Is there a way to increase the number of tries, or the timeout? The netwait= option is no good here, because the port gets closed just when the dhcprequest is initialised, so I need 5-6 secs from there. And this I guess can only be done with a higher number of tries or a higher timeout.
Lasse Riis
Udpcast mailing list Udpcast@udpcast.linux.lu https://lll.lgl.lu/mailman/listinfo/udpcast
Just as a quick addition. I just created a udpcast floppy image to take a look at it, and unfortunately, the busybox included with the floppy image is very stripped down. It doesn't contain reboot or many other option. The other versions might have a larger version of busybox.
The other option I've used is the ipmac.txt to have it get the ip addresses based on the mac address. That way to avoid the dhcp at all.
I've used busybox with g4l (ghost for linux) and added udpcast as an option.
On 28 Jun 2005 at 9:41, D Teed wrote:
Date sent: Tue, 28 Jun 2005 09:41:55 -0300 (ADT) From: D Teed dteed@artistic.ca To: Lasse Riis lists@lasseriis.dk Subject: Re: [Udpcast] Automatic reboot and DHCP-timeout Copies to: udpcast@udpcast.linux.lu
I don't know about the Cisco switch aspect, but we had a similar problem with buggy tg3 ethernet device drivers which sounds similar. Our solution was to use ipappend 1 at the head of the default file for the udpcast PXE boot. This forces it to reuse the initial DHCP address rather than request it again. The other thing that might be helpful is the command line option --rexmit-hello-interval for udp-sender. You might also want to experiment with starting the receivers or udp-sender first and see if one works better for you.
The reboot should be possible by making a udpreceiver.post shell script within your image. Busybox includes a reboot command, which may work for you.
--Donald Teed
On Tue, 28 Jun 2005, Lasse Riis wrote:
Hello
I have 2 problems with udpcast currently:
- I need to make the nodes reboot automatically efter udprcv has run.
(Otherwise I'll need a ladder to go around and reboot 36 nodes :) ) I don't know how I do this I've tried unzipping and "un-cpio-ing" the initrd but averything seems to be embedded in busybox, so I don't know what to add. Is there a "recipe" for this somewhere.
- We are using a Cisco switch here, which has an annoying habbit of
closing ports once linux releases the device just prior to a dhcp discover request. The port won't get opened till at least 5-6 seconds later. Hence udpcasts automatic dhcp-attempt fails. Is there a way to increase the number of tries, or the timeout? The netwait= option is no good here, because the port gets closed just when the dhcprequest is initialised, so I need 5-6 secs from there. And this I guess can only be done with a higher number of tries or a higher timeout.
Lasse Riis
Udpcast mailing list Udpcast@udpcast.linux.lu https://lll.lgl.lu/mailman/listinfo/udpcast
Udpcast mailing list Udpcast@udpcast.linux.lu https://lll.lgl.lu/mailman/listinfo/udpcast
+----------------------------------------------------------+ Michael D. Setzer II - Computer Science Instructor Guam Community College Computer Center mailto:mikes@kuentos.guam.net mailto:msetzerii@gmail.com http://www.guam.net/home/mikes Guam - Where America's Day Begins +----------------------------------------------------------+
http://setiathome.berkeley.edu Number of Seti Units Returned: 16,811 Processing time: 31 years, 58 days, 0 hours, 48 minutes (Total Hours: 272,953)
Busybox is just a box. Most other things in bin are simply a link to busybox. To make it run reboot, simply make a symlink to busybox called reboot, or alternately call it with 'busybox reboot'.
--Donald Teed
On Tue, 28 Jun 2005, Michael D. Setzer II wrote:
Just as a quick addition. I just created a udpcast floppy image to take a look at it, and unfortunately, the busybox included with the floppy image is very stripped down. It doesn't contain reboot or many other option. The other versions might have a larger version of busybox.
The other option I've used is the ipmac.txt to have it get the ip addresses based on the mac address. That way to avoid the dhcp at all.
I've used busybox with g4l (ghost for linux) and added udpcast as an option.
On 28 Jun 2005 at 9:41, D Teed wrote:
Date sent: Tue, 28 Jun 2005 09:41:55 -0300 (ADT) From: D Teed dteed@artistic.ca To: Lasse Riis lists@lasseriis.dk Subject: Re: [Udpcast] Automatic reboot and DHCP-timeout Copies to: udpcast@udpcast.linux.lu
I don't know about the Cisco switch aspect, but we had a similar problem with buggy tg3 ethernet device drivers which sounds similar. Our solution was to use ipappend 1 at the head of the default file for the udpcast PXE boot. This forces it to reuse the initial DHCP address rather than request it again. The other thing that might be helpful is the command line option --rexmit-hello-interval for udp-sender. You might also want to experiment with starting the receivers or udp-sender first and see if one works better for you.
The reboot should be possible by making a udpreceiver.post shell script within your image. Busybox includes a reboot command, which may work for you.
--Donald Teed
On Tue, 28 Jun 2005, Lasse Riis wrote:
Hello
I have 2 problems with udpcast currently:
- I need to make the nodes reboot automatically efter udprcv has run.
(Otherwise I'll need a ladder to go around and reboot 36 nodes :) ) I don't know how I do this I've tried unzipping and "un-cpio-ing" the initrd but averything seems to be embedded in busybox, so I don't know what to add. Is there a "recipe" for this somewhere.
- We are using a Cisco switch here, which has an annoying habbit of
closing ports once linux releases the device just prior to a dhcp discover request. The port won't get opened till at least 5-6 seconds later. Hence udpcasts automatic dhcp-attempt fails. Is there a way to increase the number of tries, or the timeout? The netwait= option is no good here, because the port gets closed just when the dhcprequest is initialised, so I need 5-6 secs from there. And this I guess can only be done with a higher number of tries or a higher timeout.
Lasse Riis
Udpcast mailing list Udpcast@udpcast.linux.lu https://lll.lgl.lu/mailman/listinfo/udpcast
Udpcast mailing list Udpcast@udpcast.linux.lu https://lll.lgl.lu/mailman/listinfo/udpcast
+----------------------------------------------------------+ Michael D. Setzer II - Computer Science Instructor Guam Community College Computer Center mailto:mikes@kuentos.guam.net mailto:msetzerii@gmail.com http://www.guam.net/home/mikes Guam - Where America's Day Begins +----------------------------------------------------------+
http://setiathome.berkeley.edu Number of Seti Units Returned: 16,811 Processing time: 31 years, 58 days, 0 hours, 48 minutes (Total Hours: 272,953)
Udpcast mailing list Udpcast@udpcast.linux.lu https://lll.lgl.lu/mailman/listinfo/udpcast
Busybox is a box, but what is in the box is determined when it is built. You can make symbolic links to busybox, but it the applet is not in the busybox it doesn't work. If you run buxybox by itself, it will show what applets are installed. You can also run these by typing buxybox command. With the floppy build busybox doesn't include reboot.
The ipmac is just a text file that contains the mac address and the ip address. If you use the web config option at the web site, you then just browser to the file, and it gets included in the build.
Here is the text from the second part of creating the image.
Upload an IP-Mac file (for automatically assigning IP addresses without a Dhcp server)? The file should be a set of lines with the Mac address to the left hand side , and the IP address at the right side Example: 00:50:56:40:42:bd 10.1.2.3 00:50:56:40:42:aa 10.1.2.4 With an IP-Mac file, you should chose UseDhcp=no, you should not enter an IP address, but you need to enter a netmask.
On 28 Jun 2005 at 10:42, D Teed wrote:
Date sent: Tue, 28 Jun 2005 10:42:29 -0300 (ADT) From: D Teed dteed@artistic.ca To: "Michael D. Setzer II" mikes@kuentos.guam.net Copies to: udpcast@udpcast.linux.lu Subject: Re: [Udpcast] Automatic reboot and DHCP-timeout
Busybox is just a box. Most other things in bin are simply a link to busybox. To make it run reboot, simply make a symlink to busybox called reboot, or alternately call it with 'busybox reboot'.
--Donald Teed
On Tue, 28 Jun 2005, Michael D. Setzer II wrote:
Just as a quick addition. I just created a udpcast floppy image to take a look at it, and unfortunately, the busybox included with the floppy image is very stripped down. It doesn't contain reboot or many other option. The other versions might have a larger version of busybox.
The other option I've used is the ipmac.txt to have it get the ip addresses based on the mac address. That way to avoid the dhcp at all.
I've used busybox with g4l (ghost for linux) and added udpcast as an option.
On 28 Jun 2005 at 9:41, D Teed wrote:
Date sent: Tue, 28 Jun 2005 09:41:55 -0300 (ADT) From: D Teed dteed@artistic.ca To: Lasse Riis lists@lasseriis.dk Subject: Re: [Udpcast] Automatic reboot and DHCP-timeout Copies to: udpcast@udpcast.linux.lu
I don't know about the Cisco switch aspect, but we had a similar problem with buggy tg3 ethernet device drivers which sounds similar. Our solution was to use ipappend 1 at the head of the default file for the udpcast PXE boot. This forces it to reuse the initial DHCP address rather than request it again. The other thing that might be helpful is the command line option --rexmit-hello-interval for udp-sender. You might also want to experiment with starting the receivers or udp-sender first and see if one works better for you.
The reboot should be possible by making a udpreceiver.post shell script within your image. Busybox includes a reboot command, which may work for you.
--Donald Teed
On Tue, 28 Jun 2005, Lasse Riis wrote:
Hello
I have 2 problems with udpcast currently:
- I need to make the nodes reboot automatically efter udprcv has run.
(Otherwise I'll need a ladder to go around and reboot 36 nodes :) ) I don't know how I do this I've tried unzipping and "un-cpio-ing" the initrd but averything seems to be embedded in busybox, so I don't know what to add. Is there a "recipe" for this somewhere.
- We are using a Cisco switch here, which has an annoying habbit of
closing ports once linux releases the device just prior to a dhcp discover request. The port won't get opened till at least 5-6 seconds later. Hence udpcasts automatic dhcp-attempt fails. Is there a way to increase the number of tries, or the timeout? The netwait= option is no good here, because the port gets closed just when the dhcprequest is initialised, so I need 5-6 secs from there. And this I guess can only be done with a higher number of tries or a higher timeout.
Lasse Riis
Udpcast mailing list Udpcast@udpcast.linux.lu https://lll.lgl.lu/mailman/listinfo/udpcast
Udpcast mailing list Udpcast@udpcast.linux.lu https://lll.lgl.lu/mailman/listinfo/udpcast
+----------------------------------------------------------+ Michael D. Setzer II - Computer Science Instructor Guam Community College Computer Center mailto:mikes@kuentos.guam.net mailto:msetzerii@gmail.com http://www.guam.net/home/mikes Guam - Where America's Day Begins +----------------------------------------------------------+
http://setiathome.berkeley.edu Number of Seti Units Returned: 16,811 Processing time: 31 years, 58 days, 0 hours, 48 minutes (Total Hours: 272,953)
Udpcast mailing list Udpcast@udpcast.linux.lu https://lll.lgl.lu/mailman/listinfo/udpcast
+----------------------------------------------------------+ Michael D. Setzer II - Computer Science Instructor Guam Community College Computer Center mailto:mikes@kuentos.guam.net mailto:msetzerii@gmail.com http://www.guam.net/home/mikes Guam - Where America's Day Begins +----------------------------------------------------------+
http://setiathome.berkeley.edu Number of Seti Units Returned: 16,811 Processing time: 31 years, 58 days, 0 hours, 48 minutes (Total Hours: 272,953)
OK, I guess in a floppy they needed to save space. You can resolve the issue by making your own busybox and check off exactly what features you want it to support. I used PXE boot so space wasn't an issue.
I've always used DHCP, so I can't provide insights on the IP Mac file.
--Donald Teed
On Wed, 29 Jun 2005, Michael D. Setzer II wrote:
Busybox is a box, but what is in the box is determined when it is built. You can make symbolic links to busybox, but it the applet is not in the busybox it doesn't work. If you run buxybox by itself, it will show what applets are installed. You can also run these by typing buxybox command. With the floppy build busybox doesn't include reboot.
The ipmac is just a text file that contains the mac address and the ip address. If you use the web config option at the web site, you then just browser to the file, and it gets included in the build.
Here is the text from the second part of creating the image.
Upload an IP-Mac file (for automatically assigning IP addresses without a Dhcp server)? The file should be a set of lines with the Mac address to the left hand side , and the IP address at the right side Example: 00:50:56:40:42:bd 10.1.2.3 00:50:56:40:42:aa 10.1.2.4 With an IP-Mac file, you should chose UseDhcp=no, you should not enter an IP address, but you need to enter a netmask.
On 28 Jun 2005 at 10:42, D Teed wrote:
Date sent: Tue, 28 Jun 2005 10:42:29 -0300 (ADT) From: D Teed dteed@artistic.ca To: "Michael D. Setzer II" mikes@kuentos.guam.net Copies to: udpcast@udpcast.linux.lu Subject: Re: [Udpcast] Automatic reboot and DHCP-timeout
Busybox is just a box. Most other things in bin are simply a link to busybox. To make it run reboot, simply make a symlink to busybox called reboot, or alternately call it with 'busybox reboot'.
--Donald Teed
On Tue, 28 Jun 2005, Michael D. Setzer II wrote:
Just as a quick addition. I just created a udpcast floppy image to take a look at it, and unfortunately, the busybox included with the floppy image is very stripped down. It doesn't contain reboot or many other option. The other versions might have a larger version of busybox.
The other option I've used is the ipmac.txt to have it get the ip addresses based on the mac address. That way to avoid the dhcp at all.
I've used busybox with g4l (ghost for linux) and added udpcast as an option.
On 28 Jun 2005 at 9:41, D Teed wrote:
Date sent: Tue, 28 Jun 2005 09:41:55 -0300 (ADT) From: D Teed dteed@artistic.ca To: Lasse Riis lists@lasseriis.dk Subject: Re: [Udpcast] Automatic reboot and DHCP-timeout Copies to: udpcast@udpcast.linux.lu
I don't know about the Cisco switch aspect, but we had a similar problem with buggy tg3 ethernet device drivers which sounds similar. Our solution was to use ipappend 1 at the head of the default file for the udpcast PXE boot. This forces it to reuse the initial DHCP address rather than request it again. The other thing that might be helpful is the command line option --rexmit-hello-interval for udp-sender. You might also want to experiment with starting the receivers or udp-sender first and see if one works better for you.
The reboot should be possible by making a udpreceiver.post shell script within your image. Busybox includes a reboot command, which may work for you.
--Donald Teed
On Tue, 28 Jun 2005, Lasse Riis wrote:
Hello
I have 2 problems with udpcast currently:
- I need to make the nodes reboot automatically efter udprcv has run.
(Otherwise I'll need a ladder to go around and reboot 36 nodes :) ) I don't know how I do this I've tried unzipping and "un-cpio-ing" the initrd but averything seems to be embedded in busybox, so I don't know what to add. Is there a "recipe" for this somewhere.
- We are using a Cisco switch here, which has an annoying habbit of
closing ports once linux releases the device just prior to a dhcp discover request. The port won't get opened till at least 5-6 seconds later. Hence udpcasts automatic dhcp-attempt fails. Is there a way to increase the number of tries, or the timeout? The netwait= option is no good here, because the port gets closed just when the dhcprequest is initialised, so I need 5-6 secs from there. And this I guess can only be done with a higher number of tries or a higher timeout.
Lasse Riis
Udpcast mailing list Udpcast@udpcast.linux.lu https://lll.lgl.lu/mailman/listinfo/udpcast
Udpcast mailing list Udpcast@udpcast.linux.lu https://lll.lgl.lu/mailman/listinfo/udpcast
+----------------------------------------------------------+ Michael D. Setzer II - Computer Science Instructor Guam Community College Computer Center mailto:mikes@kuentos.guam.net mailto:msetzerii@gmail.com http://www.guam.net/home/mikes Guam - Where America's Day Begins +----------------------------------------------------------+
http://setiathome.berkeley.edu Number of Seti Units Returned: 16,811 Processing time: 31 years, 58 days, 0 hours, 48 minutes (Total Hours: 272,953)
Udpcast mailing list Udpcast@udpcast.linux.lu https://lll.lgl.lu/mailman/listinfo/udpcast
+----------------------------------------------------------+ Michael D. Setzer II - Computer Science Instructor Guam Community College Computer Center mailto:mikes@kuentos.guam.net mailto:msetzerii@gmail.com http://www.guam.net/home/mikes Guam - Where America's Day Begins +----------------------------------------------------------+
http://setiathome.berkeley.edu Number of Seti Units Returned: 16,811 Processing time: 31 years, 58 days, 0 hours, 48 minutes (Total Hours: 272,953)
Udpcast mailing list Udpcast@udpcast.linux.lu https://lll.lgl.lu/mailman/listinfo/udpcast
D Teed wrote:
I don't know about the Cisco switch aspect, but we had a similar problem with buggy tg3 ethernet device drivers which sounds similar. Our solution was to use ipappend 1 at the head of the default file for the udpcast PXE boot. This forces it to reuse the initial DHCP address rather than request it again.
We're using the tg3 driver...so maybe this catalyst switch is not the crappiest piece of equipment in the world after all. We'll see I guess.
But I already had iappend 1 in my default file, and I tried moving it to the top (if that makes any difference) but there's no luck. So I'm going for an ipmac file. We only use dhcp for pxe-booting anyway.
The reboot should be possible by making a udpreceiver.post shell script within your image. Busybox includes a reboot command, which may work for you.
Yeah, I know about the .post script from the mailling list. But I can't figure out how to put them in the initrd (Though I've read about the makeSomething-script), and I have no idea how to --merge a new symlink?
Michael D. Setzer II wrote:
Question. How are you running udpcast? It would appear that you are not using the floppy method. If you are having problems with the dhcp, why not create an ipmac file,and have the machines get the ip addresses without even using dhcpd..
We're running it via PXE-boot only (well the sender is booted from cd-rom) I will try the ipmac file, this however I don't know where or how to put either.
On Tue, 28 Jun 2005, Lasse Riis wrote:
Yeah, I know about the .post script from the mailling list. But I can't figure out how to put them in the initrd (Though I've read about the makeSomething-script), and I have no idea how to --merge a new symlink?
Personally, I start with a ready-made image, uncompress the file, mount it with something like: mount -o loop initrd /mnt/initrd Then I copy everything into a new empty directory somewhere. (You can umount the above now.) I make the changes I want in the directory, then hop up one level and use genromfs to make a new initrd based on that directories contents. Editing within the mounted initrd can work for very small changes, but generally that can lead to space issues within the initrd. If you have troubles with the mount command, you might need to load a kernel module or recompile the kernel on the dev box where you are trying this.
--Donald Teed
D Teed wrote:
Personally, I start with a ready-made image, uncompress the file, mount it with something like: mount -o loop initrd /mnt/initrd Then I copy everything into a new empty directory somewhere. (You can umount the above now.) I make the changes I want in the directory, then hop up one level and use genromfs to make a new initrd based on that directories contents.
Well uncompressing it just yields a cpio archive, I can't really mount this. But I can use cpio to "untar" it and edit the resulting files. But I don't know how to make a new cpio archive. Is that what genromfs does?
Sorry, I didn't notice you are using floppy images - which are different than PXE boot images.
I'm not sure what makeImage is doing underneath in the case of the --cpio option, so someone else will need to answer this.
Perhaps the floppy can be accessed like this example in the genromfs man page:
mount -t romfs /dev/fd0 /mnt
--Donald Teed
On Tue, 28 Jun 2005, Lasse Riis wrote:
D Teed wrote:
Personally, I start with a ready-made image, uncompress the file, mount it with something like: mount -o loop initrd /mnt/initrd Then I copy everything into a new empty directory somewhere. (You can umount the above now.) I make the changes I want in the directory, then hop up one level and use genromfs to make a new initrd based on that directories contents.
Well uncompressing it just yields a cpio archive, I can't really mount this. But I can use cpio to "untar" it and edit the resulting files. But I don't know how to make a new cpio archive. Is that what genromfs does? _______________________________________________ Udpcast mailing list Udpcast@udpcast.linux.lu https://lll.lgl.lu/mailman/listinfo/udpcast
D Teed wrote:
Sorry, I didn't notice you are using floppy images - which are different than PXE boot images.
I'm using the PXE images actually.
But I figured it out. I bunzipped the initrd and then did: cpio -i < initrd.out
then I could replace /bin/busybox my own verison of busybox (compiled with CONFIG_REBOOT=y) and then do: genromfs -d <dir i cpio'ed to> -f initrd
That did the job It's working now. But the Cisco-switch still sucks :( But with these latest hacks I think I can make it all play nicely
Thanks for all your help
Lasse
I recall that there was one Cisco switch we used in testing that caused warnings about RX/TX flow control. We eventually learned it was something that could be managed at the switch or by the NIC, but not both. I seem to remember hearing that a piece of Cisco software on the switch could address the issue. But instead we switched to a cheap switch without any management and that enabled better performance.
I don't recall dealing with cpio before, but perhaps I did once and then did further development with basic initrd images afterwards.
--Donald Teed
On Tue, 28 Jun 2005, Lasse Riis wrote:
D Teed wrote:
Sorry, I didn't notice you are using floppy images - which are different than PXE boot images.
I'm using the PXE images actually.
But I figured it out. I bunzipped the initrd and then did: cpio -i < initrd.out
then I could replace /bin/busybox my own verison of busybox (compiled with CONFIG_REBOOT=y) and then do: genromfs -d <dir i cpio'ed to> -f initrd
That did the job It's working now. But the Cisco-switch still sucks :( But with these latest hacks I think I can make it all play nicely
Thanks for all your help
Lasse
Udpcast mailing list Udpcast@udpcast.linux.lu https://lll.lgl.lu/mailman/listinfo/udpcast
-
D Teed -
Lasse Riis -
Michael D. Setzer II