"RUNNING HANDLER [handlers : wait for weave to listen]" timeout
NinoFloris opened this issue · comments
I have tried to get this great starting point for a cluster online but I get stuck with Weave not getting online correctly.
This is the console output leading up to the timeout of the ansible step.
https://gist.github.com/NinoFloris/89609002e852ce87fb3470320202e215
I don't exactly know how to debug this, Ssh'ing into a host and running docker ps -a gives this result
CONTAINER ID IMAGE COMMAND CREATED STATUS PORTS NAMES
b8e1aa9af01c weaveworks/weaveexec:1.4.4 "/bin/false" 2 minutes ago Created weavevolumes-1.4.4
So at least docker is running fine but systemctl shows weave and weaveproxy as dead
weave.service loaded inactive dead Weave Network Router
weaveproxy.service loaded inactive dead Weave proxy for Docker API
Manually starting weaveproxy.service shows weave as 'activating' not 'active' (is this because of docker attach that keeps the service connected in the foreground?).
After starting weaveproxy.service my docker ps looks like this
CONTAINER ID IMAGE COMMAND CREATED STATUS PORTS NAMES
37b410f490d7 weaveworks/weaveexec:1.4.4 "/home/weave/sigproxy" 20 seconds ago Up 19 seconds amazing_stallman
2406bda553a0 weaveworks/weave:1.4.4 "/home/weave/weaver -" 21 seconds ago Up 20 seconds weave
885176cfe20e weaveworks/weaveexec:1.4.4 "/home/weave/weavepro" 21 seconds ago Up 21 seconds weaveproxy
b8e1aa9af01c weaveworks/weaveexec:1.4.4 "/bin/false" 7 minutes ago Created weavevolumes-1.4.4
I would love some pointers as where to look because this is quite new terrain for me, the project is great though. I had almost started working on something exactly like this.
EDIT: This is on DigitalOcean
@NinoFloris if you export ANSIBLE_LOG="-vvvv" before running ansible again it should give you good debug output
I don't really see anything out of the ordinary here...
https://gist.github.com/NinoFloris/5485091483ca132120fce3169ecdac62
Maybe what I would like to know more is if there was a recent sucessful deployment by Capgemini on digitalocean. As it is stated to be project level supported in the docs it should just work right?
@NinoFloris I just tested the master branch on digitalocean and it worked for me. I did notice that the default master_instance_type and slave_instance_type of 512mb isnt enough and I have seen memory issues in some of the software used in the project. I usually use at least the 2GB for master and 4GB for slaves.
Ahhh wow, I would have never thought of that angle. I will try this tomorrow. Many thanks!
Would a contribution of better docs be something I could help with once I get a cluster up and running and have a better feeling for all the components involved?
Yes please that would be awesome!
It's strange and I don't like that I cannot solve this problem on my own but I still cannot get past this ansible step.
What my exact setup is:
- TF_VAR for region (ams2) but I have also tried the default, lon1.
- TF_VAR for master and slave instance types, as you suggested 2gb and 4gb.
- APOLLO_PROVIDER set to digitalocean.
- TF_VAR_do_token set to a valid token.
I have had runs with TF_VAR_public_key_file set to a correct path and runs where I did not export it as it does not seem to do much?
I see the droplets come up in the DO web interface with the correct region and size.
I have had ANSIBLE_LOG=-vvvv but this did not reveal any insights on this issue.
My Terraform version is v0.6.14, Ansible is 2.0.1.0, and I run this on a normal OSX 10.11
Looking at ref master in my .git directory, the Apollo version I use is commit hash 5ec6257 e.g. current HEAD (5 april 2016)
Are there any variables that I have not set correctly? As this is all that I could find in the docs and code that really required input.
@NinoFloris you are correct, I started getting the same error. Only happens on the first run though. Seems to be a repeat of this #662. When I run ./bootstrap/apollo-launch.sh ansible_playbook_run the second time everything comes up fine.
labelling as bug - think we need to look deeper into this and provide a fix.
Not sure if this is helpful info, but I'm getting the same problem running the vagrant build (using master branch)