distcc (dcc_connect_by_name_real) failed to resolve hostname

[LIsLinuxIsSogood]

I am wondering if someone could help explain the error. Distcc I thought had several options for using addressing. Why is this one option failing?

[krinn]

Dunno what you did, either in kernel makefile or distcc/hosts but twice time in your output it not only show the extra ' but also an extra = after the IP address
And of course, it couldn't resolve it.

To set how distcc works outside portage, you have /etc/distcc/hosts to setup hosts and /etc/env.d/02distcc to setup distcc parameters.
You can use make CC=distcc -jX to use distcc with your kernel, no need to set hosts in the kernel make file

[Hu]

Please show us how you configured the list of hosts to which distcc should distribute.

[szatox]

[LIsLinuxIsSogood]

Not a typo in the Makefile, I've checked numerous times. It is somewhat puzzling, and I'm confused...going to try with krinn solution to make CC=distcc -jN and see if that works.

Hu, I don't know if I mentioned it bust distcc works perfectly ok otherwise. The hosts are working to distribute other packages, e.g. through portage or whatever.

[LIsLinuxIsSogood]

I removed all changes to Makefile, those hostname errors are gone. Compiling the kernel with the mentioned settings is working OK, but for whatever reason hardly anything gets distributed to the other machine. Most likely something to do with the size of the code though, so I guess in theory it works now. Just trying to find a better way to compile the kernel than locally by using the other PC's faster processor.

Update: Nevermind, I see there are some portions the kernel is processing the files into .o files automatically and not distributing those For example, compiling the kernel with CC=distcc, sends just a handful of files for compilation to the host machines (less than 10 files..which out of 100's of files isn't substantial) I want to save time so I think running the entire compilation on the host machine (using ftp or scp to copy the file there and then run make from there).

[LIsLinuxIsSogood]

With distcc

real 25m54.869s
user 27m50.162s
sys 1m56.539s

without distcc

real 26m23.540s
user 28m8.482s
sys 1m54.629s


The small difference amounts to probably a dozen or so files that were compiled on the host machine.

Does anyone know why the kernel won't make use of distcc, is it because there are numerous other compiles or compiling procedures (for C language stuff)?

EDIT: With -j4 reduced real time on local compile to
real 11m49.024s
user 41m48.988s
sys 2m43.052s

[LIsLinuxIsSogood]

Here is my hosts file for distcc:

[LIsLinuxIsSogood]

Here's another debug set of info when compiling kernel...with distcc

Oct 11 20:50:03 [distccd] (dcc_listen_by_addr) listening on 192.168.1.5:3632
Oct 11 20:50:03 [distccd] (dcc_standalone_server) 4 CPUs online on this server
Oct 11 20:50:03 [distccd] (dcc_standalone_server) allowing up to 6 active jobs
Oct 11 20:50:03 [distccd] (dcc_log_daemon_started) preforking daemon started (3.2rc1 x86_64-pc-linux-gnu, built Sep 5 2017 07:54:44)

So it looks like everything working, but almost every aspect of the kernel gets compiled locally regardless!

[krinn]

It's common to have options set with 0 disable, 1 enable, those just mean false/true
DISTCC_FALLBACK and DISTCC_SKIP_LOCAL_RETRY are similar, both handle how distcc react to a problem and if distcc bypass the problem by building the file on localhost or not.

DISTCC_FALLBACK:
For me it should be on, distcc fail to distribute to host Z, and rebuild the job locally.
People who want this off are distcc devs for debugging, or users that use two hosts, and one host is a critical helper. ie: people using a big cpu to help their arm building a package, if "big cpu" host cannot take the job, something is wrong, and it might be real better to just stop building and fix the issue than letting the arm cpu continue to build everything on its own.
DISTCC_SKIP_LOCAL_RETRY:
It should just always be off: if the host goes offline, it will fail to compile and distcc will build the file locally, because host is now offline, next attempt will be handle by DISTCC_FALLBACK (because next job will be attempt to be distribute to it).
But it might be a great help to someone using really a lot of hosts that this option is on, in order to help you identify a problem in the farm.

For me best settings that should be use are DISTCC_FALLBACK=1 DISTCC_SKIP_LOCAL_RETRY=0 (i think those are default settings in gentoo).
If you don't have a farm of hosts and use one host to help another one that is really really weak, DISTCC_FALLBACK=0 looks better.


For performance, because you use distcc to build a kernel, you will have better performance from building the kernel directly on the fast host instead of using distcc.