Re: [gentoo-user] distcc configuration
michael higgins wrote:
On Thu, 16 Jun 2005 14:08:46 -0700
Zac Medico [EMAIL PROTECTED] wrote:
michael higgins wrote:
I have a couple of questions about using distcc. I have two machines. One is
significantly faster than the other, both x86 (pentium 2 and amd athlon-xp).
First, a big thanks to all who replied.
I've followed the http://www.gentoo.org/doc/en/distcc.xml instructions for
configuring portage to use distcc. Should I need to continue with the
instructions for working with automake?
Portage does this automatically as long as you have the corresponding
FEATURES enabled.
This is what I was hoping to hear.
Also, how do I tell the faster machine to just use it's own power and not
attempt to use the slower one?
When you run distcc-config --set-hosts on the faster machine you want to
exclude the slower machine.
I tried so far, getting distccd running on both machines, each distcc-config
--set-hosts has one IP entry, that of the other machine. Is this right?
You may want to include localhost.
Interesting... I wonder why?
Well, sometimes you may not want to include localhost. Maybe it has enough
load already.
I ran distcc-gnome and saw no activity reported on the faster machine when
emerging something on the slower one. So, it would seem it's not working.
Can anyone give me some hints? The manpage for distcc doesn't seem (to me)
to be much related to the gentoo how-to... '-)
TIA,
#!/bin/bash
source /etc/make.globals
source /etc/make.conf
export DISTCC_DIR=${PORTAGE_TMPDIR}/portage/.distcc
exec /usr/bin/distccmon-gnome
So, I just installed a package on the faster machine. It did try to use
another, but found none and compiled locally, but spit out an error. Maybe
adding localhost will fix this?
If the faster machine isn't going to use any distcc nodes (other than
localhost) then you should remove distcc from FEATURES.
I tried emerging the same package on the slower machine and running this
script above. Nothing came up in it. However, I noticed that each time the
compiler went to run something, there was network activity. 'top' on the
faster machine showed distccd working and launching the compiler, afaict.
You can enable logging in /etc/conf.d/distccd.
So, I have to wonder if/why the monitors don't work for me...
export DISTCC_DIR=${PORTAGE_TMPDIR}/portage/.distcc works for me. Maybe it
will help if you use lsof to find out what files distcc has open.
Next time I'll just set the debug level and log location to see what actually
happened.
Thanks again, foax.
--
gentoo-user@gentoo.org mailing list
Re: [gentoo-user] distcc configuration
On Thu, 16 Jun 2005 14:08:46 -0700
Zac Medico [EMAIL PROTECTED] wrote:
michael higgins wrote:
I have a couple of questions about using distcc. I have two machines. One
is significantly faster than the other, both x86 (pentium 2 and amd
athlon-xp).
First, a big thanks to all who replied.
I've followed the http://www.gentoo.org/doc/en/distcc.xml instructions for
configuring portage to use distcc. Should I need to continue with the
instructions for working with automake?
Portage does this automatically as long as you have the corresponding
FEATURES enabled.
This is what I was hoping to hear.
Also, how do I tell the faster machine to just use it's own power and not
attempt to use the slower one?
When you run distcc-config --set-hosts on the faster machine you want to
exclude the slower machine.
I tried so far, getting distccd running on both machines, each
distcc-config --set-hosts has one IP entry, that of the other machine. Is
this right?
You may want to include localhost.
Interesting... I wonder why?
I ran distcc-gnome and saw no activity reported on the faster machine when
emerging something on the slower one. So, it would seem it's not working.
Can anyone give me some hints? The manpage for distcc doesn't seem (to me)
to be much related to the gentoo how-to... '-)
TIA,
#!/bin/bash
source /etc/make.globals
source /etc/make.conf
export DISTCC_DIR=${PORTAGE_TMPDIR}/portage/.distcc
exec /usr/bin/distccmon-gnome
So, I just installed a package on the faster machine. It did try to use
another, but found none and compiled locally, but spit out an error. Maybe
adding localhost will fix this?
I tried emerging the same package on the slower machine and running this script
above. Nothing came up in it. However, I noticed that each time the compiler
went to run something, there was network activity. 'top' on the faster machine
showed distccd working and launching the compiler, afaict.
So, I have to wonder if/why the monitors don't work for me...
Next time I'll just set the debug level and log location to see what actually
happened.
Thanks again, foax.
--
|\ /| | | ~ ~
| \/ | |---| `|` ?
||ichael | |iggins\^ /
michael[dot]higgins[at]snet[dot]net
--
gentoo-user@gentoo.org mailing list
RE: [gentoo-user] distcc configuration
There were some changes to the tool chain which means that distcc wont work across different x86 archs (e.g., athlon and pentium) and the devs wont change it (marked as WONT FIX on the bug). Hmm, I haven't heard anything about this and have been using distcc on an x86 athlon network. Each host seems to be doing the write thing as all of the machines are recently sync'd and built. What exactly were the nature of the changes and how does this introduce a bug? Also if it is truly a bug, why does the handbook still include the references to building using distcc? Wouldn't failures in distcc lead to more installation problems? The apparent stability of a gentoo-based system would then be subject to question, not something I'd like to see happen... -- gentoo-user@gentoo.org mailing list
RE: [gentoo-user] distcc configuration
read the bug :) - the bug numbers in the email, or search the forums like I did. Are you sure its building? distcc connects to the machine in question, but doesnt actually build anything as it cant run gcc as the toolchain does not call it gcc internally anymore, but includes i[3456]86 in the naming scheme. So if you are crossing one of these boundaries (i.e., trying o build for a K6 which is i586 using a later pentium which is i686), it wont work. I see I was wrong in citing athlon/pentium as they are both i686 - that being said my athlon pentium combination is not working at the moment, which I will have to look into again soon. BillK On Fri, 2005-06-17 at 09:04 -0400, Dave Nebinger wrote: There were some changes to the tool chain which means that distcc wont work across different x86 archs (e.g., athlon and pentium) and the devs wont change it (marked as WONT FIX on the bug). Hmm, I haven't heard anything about this and have been using distcc on an x86 athlon network. Each host seems to be doing the write thing as all of the machines are recently sync'd and built. What exactly were the nature of the changes and how does this introduce a bug? Also if it is truly a bug, why does the handbook still include the references to building using distcc? Wouldn't failures in distcc lead to more installation problems? The apparent stability of a gentoo-based system would then be subject to question, not something I'd like to see happen... -- William Kenworthy [EMAIL PROTECTED] Home! -- gentoo-user@gentoo.org mailing list
RE: [gentoo-user] distcc configuration
read the bug :) - the bug numbers in the email, or search the forums like I did. What a pain. I'm sure the change was meant to solve another problem, but it's unfortunate that it broke distcc in the process. Are you sure its building? distcc connects to the machine in question, but doesnt actually build anything as it cant run gcc as the toolchain does not call it gcc internally anymore, but includes i[3456]86 in the naming scheme. So if you are crossing one of these boundaries (i.e., trying o build for a K6 which is i586 using a later pentium which is i686), it wont work. I see I was wrong in citing athlon/pentium as they are both i686 - that being said my athlon pentium combination is not working at the moment, which I will have to look into again soon. Yeah, I'm building, but you're right in that all participating hosts are using the i686 gcc. This would explain why it's been working for me all of this time... -- gentoo-user@gentoo.org mailing list
Re: [gentoo-user] distcc configuration
michael higgins wrote:
I have a couple of questions about using distcc. I have two machines. One is
significantly faster than the other, both x86 (pentium 2 and amd athlon-xp).
I've followed the http://www.gentoo.org/doc/en/distcc.xml instructions for
configuring portage to use distcc. Should I need to continue with the
instructions for working with automake?
Portage does this automatically as long as you have the corresponding FEATURES
enabled.
Also, how do I tell the faster machine to just use it's own power and not
attempt to use the slower one?
When you run distcc-config --set-hosts on the faster machine you want to
exclude the slower machine.
I tried so far, getting distccd running on both machines, each distcc-config
--set-hosts has one IP entry, that of the other machine. Is this right?
You may want to include localhost.
I ran distcc-gnome and saw no activity reported on the faster machine when
emerging something on the slower one. So, it would seem it's not working.
Can anyone give me some hints? The manpage for distcc doesn't seem (to me) to
be much related to the gentoo how-to... '-)
TIA,
#!/bin/bash
source /etc/make.globals
source /etc/make.conf
export DISTCC_DIR=${PORTAGE_TMPDIR}/portage/.distcc
exec /usr/bin/distccmon-gnome
Zac
--
gentoo-user@gentoo.org mailing list
Re: [gentoo-user] distcc configuration
Hi, In order for distcc to actually work there are a few steps to make: If you have one machine 192.168.0.1 second 192.168.0.2 and assuming 192.168.0.1 is the power These are the things you should do on the slow machine (there may be more, i've done it some time ago :)): first in /etc/make.conf make sure MAKEOPTS is something like this MAKEOPTS=-jN where N is the number of CPUs per machine + 2 (1 for each machine). This makes sure make will try to use as many parallel processes as it should make sure on both machines you have set-up distccd to listen and to accept connections, check /etc/conf.d/distcc on 192.168.0.2 make # distcc-config --set-hosts 192.168.0.1:3632 3 192.168.0.2:3632 2 this ensures the slow machine will first try the fast one and then then itself. The number 3 and 2 assume the first machine is a dual processor and the second is a single processor (number of cpus + 1 per machine) on 192.168.0.1 make # distcc-config --set-hosts 192.168.0.1:3632 3 only the power machine is used this way or add the same as above if u also want the slow in the portage features in /etc/make.conf there should be distcc Of course, both machine should use the same build environment (gcc + glibc + binutils) That's about it. # ps aux from time to time on the machine will show some gcc process and you will know it works :) If it works once it will work always. Hope this helps, Catalin michael higgins wrote: I have a couple of questions about using distcc. I have two machines. One is significantly faster than the other, both x86 (pentium 2 and amd athlon-xp). I've followed the http://www.gentoo.org/doc/en/distcc.xml instructions for configuring portage to use distcc. Should I need to continue with the instructions for working with automake? Also, how do I tell the faster machine to just use it's own power and not attempt to use the slower one? I tried so far, getting distccd running on both machines, each distcc-config --set-hosts has one IP entry, that of the other machine. Is this right? I ran distcc-gnome and saw no activity reported on the faster machine when emerging something on the slower one. So, it would seem it's not working. Can anyone give me some hints? The manpage for distcc doesn't seem (to me) to be much related to the gentoo how-to... '-) TIA, -- gentoo-user@gentoo.org mailing list
Re: [gentoo-user] distcc configuration
There were some changes to the tool chain which means that distcc wont work across different x86 archs (e.g., athlon and pentium) and the devs wont change it (marked as WONT FIX on the bug). The workaround is to add CC-gcc and CXX=g++ to make .conf on the client, however this may break other things, not to mention that it isnt currently working for me with gcc3.4 (multiple toolchain scenario). This is a real pita as distcc is most useful when using a fast modern machine supporting slow older machines, which usually means different x86 archs. If you use distcc, turn logging on on the server and make sure its really working, and not going through the motions and silently erroring off. see the distcc buglist (#84942) and I think some forum posts for the gory details ... BillK On Thu, 2005-06-16 at 12:56 -0400, michael higgins wrote: I have a couple of questions about using distcc. I have two machines. One is significantly faster than the other, both x86 (pentium 2 and amd athlon-xp). I've followed the http://www.gentoo.org/doc/en/distcc.xml instructions for configuring portage to use distcc. Should I need to continue with the instructions for working with automake? Also, how do I tell the faster machine to just use it's own power and not attempt to use the slower one? I tried so far, getting distccd running on both machines, each distcc-config --set-hosts has one IP entry, that of the other machine. Is this right? I ran distcc-gnome and saw no activity reported on the faster machine when emerging something on the slower one. So, it would seem it's not working. Can anyone give me some hints? The manpage for distcc doesn't seem (to me) to be much related to the gentoo how-to... '-) TIA, -- |\ /| | | ~ ~ | \/ | |---| `|` ? ||ichael | |iggins\^ / michael[dot]higgins[at]snet[dot]net -- William Kenworthy [EMAIL PROTECTED] Home! -- gentoo-user@gentoo.org mailing list
