#weavenetwork

VM or physical?

VM

the above takes 170ms on my laptop.

it's an Azure VM

what can possibly slow down so much ethtool?

not ethtool itself, but docker container execution.

trying with a different vm size

maybe it can help

even worse :)

$ sudo time docker run --privileged --net=host zettio/weavetools:0.8.0 /bin/ethtool -K eth0 tx on

0.01user 0.01system 0:03.31elapsed 1%CPU (0avgtext+0avgdata 14080maxresident)k

0inputs+0outputs (0major+844minor)pagefaults 0swaps

nothing more I can do here right now. changing the timeout requires publishing a new release.

yaronr: ping

moonfish1: do you think an older weave version could help?

yes, 0.7.0 would work, but it's going to be tricky to get that downloaded since version-alignment was only introduced in 0.8.0.

(getting the 0.7.0 script is easy, getting it to use the 0.7.0 images isn't)

andreaturli: I am astonished that docker is so slow on those azure machines.

moonfish: mmm, not sure it is true for all the containers

really? you have some containers that run significantly faster?

let me try

moonfish: that is a little better, but not good

moonfish: http://stackoverflow.com/questions/25963729/why...

this seems to help me as well

andreautil: that first line uses "-d" which daemonises the container, so it doesn't actually wait for it to start properly.

re slow build... weave isn't building containers on the fly, so I don't think that is an issue.

moonfish: thanks again for your help

I'm now using a new instance on Azure, same ubuntu image, and now weave launch works fine

:S

wow. that is weird.

I havent touched anything

I simply can notice that network is slightly faster ...

but nothing more

$ sudo time docker run --privileged --net=host zettio/weavetools:0.8.0 /bin/ethtool -K eth0 tx on

Cannot change tx-checksumming

Could not change any device features

Command exited with non-zero status 1

0.01user 0.01system 0:00.47elapsed 5%CPU (0avgtext+0avgdata 8636maxresident)k

0inputs+0outputs (0major+2095minor)pagefaults 0swaps

this seems a reasonable time, not super, but still

yes.

what scares me is that I still don't know the problem, if any

well, VM performance in public clouds can vary massively.

yeah, but work vs doesn't it is too much!

anyway thanks again for your time and great help!

well, that's purely due to the hard-coded 5s timeout in the weave container process start. which we will do something about.

Hi, I'm trying to run a docker container using weave on a remote docker host .... yeah I know .....

docker run -v /var/run/docker.sock:/var/run/docker.sock -v /proc:/hostproc -v /tmp:/tmp --privileged=true --rm --env PROCFS=/hostproc --net=host -ti igneoussystems/weave:latest 10.0.1.5/20 -d --name=Topo.slab0.lithos5942550.0 --privileged=true --with-dns --dns-search=weave.local lithos:dev /bin/sh -c ping -c 3 plume.weave.local"

was trying to check the dns works here .... problem is it seem docker arg parsing gets confused

it says : flag provided but not defined: --with-dns

How have you guys approached the problem of exposing a weave subnet to hosts that don't have weave running? I've attempted but thus far failed to route through a machine that does have weave running and has exposed the weave network locally. Is there anything that makes such a setup impossible? I think I'm doing something wrong, but would like to know that I'm not trying something that would be obviously impossible.

diego_diego: it's almost certainly possible, but exactly what needs doing, and the complexity thereof very much depends on the host configuration and what is between them. So there really is no generic answer. Except "just run weave everywhere", of course ;)

heh, gotcha thanks moonfish