nlf / dhyve-os Goto Github PK
View Code? Open in Web Editor NEWa tiny OS for running docker in xhyve
a tiny OS for running docker in xhyve
I'm wondering why each of those files is archived by tar instead of a plain text.
i don't like how it is now, it needs to be better. i hate that i have to run make build
manually if i change files and then make
again to get the artifacts out of it.
maybe it's best to just build everything in the image and then create a container without running it to extract the artifacts?
Because dhyve uses 8.8.8.8 for DNS the VM is unable to resolve DNS entries that the host can via VPN. Please offer a way to use the host for DNS resolution.
I tried build dhyve-os to support our enterprise but it has been running now for over 55 mins is this normal?
it's seems stuck here...
linux 4.3.3 Configuring
linux 4.3.3 Building
...
finally it continued then I got this and it failed:
cc1: some warnings being treated as errors
scripts/Makefile.modpost:113: recipe for target 'net/ipv4/netfilter/arptable_filter.mod.o' failed
make[2]: *** [net/ipv4/netfilter/arptable_filter.mod.o] Error 1
make[2]: *** Waiting for unfinished jobs....
Makefile:1095: recipe for target 'modules' failed
make[1]: *** [modules] Error 2
package/pkg-generic.mk:196: recipe for target '/tmp/buildroot/output/build/linux-4.3.3/.stamp_built' failed
make: *** [/tmp/buildroot/output/build/linux-4.3.3/.stamp_built] Error 2
make: *** [build] Error 2
most users seem to just want the latest docker version, so let's preinstall whatever the latest is and keep the option to the init script in place so people can change it out easily
These lines in the Makefile:
$(eval STR_CREATED=$$(shell docker inspect -f '{{.Created}}' $(BUILD_IMAGE) 2>/dev/null))
$(eval IMG_CREATED=$$(shell date -j -u -f "%F %T" "$$(STR_CREATED)" +"%s" 2>/dev/null \
|| echo 0))
are supposed to get the created timestamp of the image. But on my system, docker inspect
returns an ISO 8601 time stamp like 2015-12-24T08:18:59.927830094Z
, and that doesn't match %F %T
. (%FT%T
works.)
Does %F %T
work for others? Is the problem specific to my Docker version?
$ docker inspect -f '{{.Created}}' dhyve-os-builder
2015-12-24T08:18:59.927830094Z
$ date -j -u -f "%F %T" "2015-12-24T08:18:59.927830094Z" +"%s"
Failed conversion of ``2015-12-24T08:18:59.927830094Z'' using format ``%F %T''
date: illegal time format
usage: date [-jnu] [-d dst] [-r seconds] [-t west] [-v[+|-]val[ymwdHMS]] ...
[-f fmt date | [[[mm]dd]HH]MM[[cc]yy][.ss]] [+format]
$ date -j -u -f "%FT%T" "2015-12-24T08:18:59.927830094Z" +"%s"
Warning: Ignoring 11 extraneous characters in date string (.927830094Z)
1450945139
$ docker version
Client:
Version: 1.8.2
API version: 1.20
Go version: go1.4.2
Git commit: 0a8c2e3
Built: Thu Sep 10 19:10:10 UTC 2015
OS/Arch: darwin/amd64
Server:
Version: 1.9.1
API version: 1.21
Go version: go1.4.3
Git commit: a34a1d5
Built: Fri Nov 20 17:56:04 UTC 2015
OS/Arch: linux/amd64
It seems the rootfs/etc/init.d/S51docker
script is overwriting the docker binary on startup with a 0-byte file.
Hey,
getting docker from https://get.docker.io/builds/Linux/x86_64/docker-${docker_version}
no longer works. It throws a 403. If you add .tgz
on the end however it works. You can run this through gunzip then tar to get the binaries.
Since the overlay size wasn't big enough to hold all the files. I put docker
in /usr/bin/
and docker-*
in /bin/
Cheers,
Ben
Install docker with an init script which reads an env variable (docker_version=1.10.0
) and installs the appropriate version.
The default docker Debian package sets
ulimit -n 1048576
if [ "$BASH" ]; then
ulimit -u 1048576
else
ulimit -p 1048576
fi
DhyveOS has much smaller limits, and although they can be changed using --ulimit
when doing docker run
, this causes some images that work using the Debian docker to fail with DyhveOS. It would be cool if these limits could be increased.
Currently docker will bind ports to 0.0.0.0 on the VM which means that when running docker-compose port postgres 5432
(or just using docker inspect) we get back 0.0.0.0:<bound port>
. This means that I need to replace the 0.0.0.0
with the VM's IP in order to use it, putting back some of the pain that DLite might otherwise save.
It would be very useful if --ip=192.168.x.x
was passed to the docker daemon when starting so that it would by default bind ports to the interface that we can reach from outside the VM. This way docker inspect and docker-compose would show an IP:PORT
combination that can be connected to as-is.
That way I never have to care what the IP address of the VM is.
I've been using -o noatime,discard,ssd,autodefrag,compress=lzo,space_cache
when mounting btrfs for docker storage on ec2. But we should do some tests to see what's good under xhyve.
I see that 3.0.0 has been replaced a few times now, can we bump the version each time so I know if I'm out of date?
Sorry to be a pest, lol.
Ty,
Trevor
Hi!
This is related to dlite (v1 with NFS) and most likely future versions of dlite/dhyve-os since I've read that you are planning on moving back to NFS from 9P. Really looking forward to a future v2 of dlite with NFS since the performance of 9P is really terrible for my use cases.
The only standing issue that I've had with dlite is the annoying and frequent behavior that file updates are not synced between server and host unless you wait for 5-10 seconds.
After the finding this stack overflow question and including the actimeo=1
option in the mount command in the legacy branch everything seems to be working perfectly.
Is this something you would consider including in the upcoming NFS solution for v2?
In c61d125 CONFIG_FUSE_FS
was unset.
https://github.com/nlf/dhyve-os/blob/master/config/kernel#L2279
Are you open to a PR to bring FUSE support back? e.g
diff --git a/config/kernel b/config/kernel
index 818ed0b..7fb71f9 100644
--- a/config/kernel
+++ b/config/kernel
@@ -2253,7 +2253,8 @@ CONFIG_FANOTIFY=y
# CONFIG_QUOTA is not set
# CONFIG_QUOTACTL is not set
# CONFIG_AUTOFS4_FS is not set
-# CONFIG_FUSE_FS is not set
+CONFIG_FUSE_FS=y
CONFIG_OVERLAY_FS=y
#
# Caches
(I haven't preemptively sent a PR because I'm still using 2.2.0 and dlite 1.1.4, so haven't built a 3.0.0 beta to test CONFIG_FUSE_FS
)
dhyve-os currently uses the overlay storage driver, which has a known issue that causes yum
to fail in containers. (Documented in https://docs.docker.com/engine/userguide/storagedriver/overlayfs-driver/)
Would it be reasonable to support and default to a different storage driver, such as AUFS or devicemapper?
Discussion directly coming from here: nlf/dlite#61
To be future-proof, a build and test environment should be created, that is able to build multiple images for each (?!) minor docker version?
My proposal would be:
Travis CI
build an image matrix and test it with the current version of dlite
and docker
Should I try to implement such a thing with travis?
should remove the current resolv.conf in favor of this
Hey there. I tried the 2.3.1 image from my PR, but got a kernel error :( :
❯ dlite stop
❯ dlite update -v 2.3.1
❯ dlite start
❯ ps aux | grep dlite
leipert 6470 0.0 0.0 2425204 428 s002 R+ 10:35AM 0:00.00 grep --color=auto dlite
❯ sudo dlite daemon
COM1 connected to /dev/ttys003
kexec: failed to load kernel /Users/leipert/.dlite/bzImage
I also tried to dlite install
with -v 2.3.1
, which doesn't work either.
The 2.3.0 image starts, but they changed the API version from 1.11.0
to 1.11.1
think this is due to a kernel configuration problem
A declarative, efficient, and flexible JavaScript library for building user interfaces.
🖖 Vue.js is a progressive, incrementally-adoptable JavaScript framework for building UI on the web.
TypeScript is a superset of JavaScript that compiles to clean JavaScript output.
An Open Source Machine Learning Framework for Everyone
The Web framework for perfectionists with deadlines.
A PHP framework for web artisans
Bring data to life with SVG, Canvas and HTML. 📊📈🎉
JavaScript (JS) is a lightweight interpreted programming language with first-class functions.
Some thing interesting about web. New door for the world.
A server is a program made to process requests and deliver data to clients.
Machine learning is a way of modeling and interpreting data that allows a piece of software to respond intelligently.
Some thing interesting about visualization, use data art
Some thing interesting about game, make everyone happy.
We are working to build community through open source technology. NB: members must have two-factor auth.
Open source projects and samples from Microsoft.
Google ❤️ Open Source for everyone.
Alibaba Open Source for everyone
Data-Driven Documents codes.
China tencent open source team.