Need pointers to get Lucida working post docker image creation

364 views
Skip to first unread message

Lokender Kaushik

unread,
Sep 18, 2016, 3:01:25 AM9/18/16
to Lucida Users
My progress so far:

My Dev env:
  1. Macbook Air osx (Yosemite - 10.10.5), 8 GB 1600 MHz DDR3, Graphics - Intel HD Graphics 6000 1536 MB
My progress so far:
  1. Cloned the lucida repository
  2. Installed docker and ran "build -t my-lucida" - It created a 16 gb image

My questions:
  1. Being on osx - Can I use docker to run lucida? Will it work with my machine specs? What other options do I have?
  2. How to proceed after the image is created. I tried:
    1. Running lucida/tools/start_all.sh inside the container. It gave me errors like "

      Starting service in /usr/local/lucida/tools/../lucida/commandcenter

      gnome-terminal -x bash -c "source /usr/local/lucida/lucida/commandcenter/../../tools/python_2_7_9/bin/activate && python app.py ; read -n1"

      /bin/sh: gnome-terminal: command not found

      make: *** [start_server] Error 127

      Starting service in /usr/local/lucida/tools/../lucida/speechrecognition/kaldi_gstreamer_asr

      ./simple_start.sh

      ./simple_start.sh: line 3: gnome-terminal: command not found

      make: *** [start_server] Error 127"
    2. I also tried downloading 'https://github.com/jhauswald/lucida/blob/master/docker-compose.yml' and running it from lucida home. It gives me errors "

      docker-compose up

      Creating lucida_lucida_services_1

      Creating lucida_lucida_asr_1


      ERROR: for lucida_asr  Cannot start service lucida_asr: Cannot link to a non running container: /lucida_lucida_services_1 AS /lucida_lucida_asr_1/lucida_lucida_services_1

      ERROR: Encountered errors while bringing up the project."
P.S. - I am somewhat new to Docker as well as Lucida.

Yunsheng Bai

unread,
Sep 18, 2016, 10:47:56 AM9/18/16
to Lucida Users
Hi,

Running Lucida inside the docker container is not recommended, which is why "start_all.sh" failed, because you cannot use "gnome-terminal" inside the docker container. I can provide two options:

1. Install a Linux virtual machine. Personally, I use Parallel Desktop, but you can try VirtualBox, VMware, etc. The advantages are that you can develop and test more easily, and "gnome-terminal" is available in your VM. One disadvantage is that installing the virtual machine isn't cost-free.

2. Pull the official repo from https://github.com/claritylab/lucida (different from the repo you are currently using), and follow the instructions under the section "Lucida Docker Deployment". You will be using Kubernetes rather than docker compose, and pulling an image from our docker hub repo https://hub.docker.com/r/claritylab/lucida/tags/ rather than build your own. The advantages are that the steps are relatively simple, and take less amount of time. One disadvantage is that developing is harder inside the docker container.

Let me know what you think and report any problems you have. Thanks a lot! 

Chris Pitchford

unread,
Sep 26, 2016, 7:05:37 PM9/26/16
to Lucida Users
This sounds like sage advice! I'm trying it now for a demo. No development (just troubleshooting, if needed) at this stage.

Wish me luck!

Chris Pitchford

unread,
Sep 27, 2016, 10:55:49 AM9/27/16
to Lucida Users
Hello! Trying the instructions at: https://github.com/claritylab/lucida/tree/master/tools/deploy

I pulled v2.1 from Docker hub. Then, per the instructions, `sudo ./cluster_up.sh`

But the instructions refer to `sudo ./start_services.sh` and I get repeated errors: `kubectl: command not found`

When I run `docker ps -a` I get:

5685e04d55b7        gcr.io/google_containers/hyperkube-amd64:v1.3.0   "/hyperkube kubelet -"   16 hours ago


On Sunday, September 18, 2016 at 8:47:56 AM UTC-6, Yunsheng Bai wrote:

Yunsheng Bai

unread,
Sep 27, 2016, 11:18:17 AM9/27/16
to Lucida Users
Hi Chris,

There should be multiple docker containers running when you type  `docker ps -a`. Try stopping and removing all the running containers (perhaps only 1 according to your description) and starting the cluster again with `sudo ./cluster_up.sh`. Fix any permission error if exists. Make sure your docker engine is updated and does not prevent you from starting multiple docker containers. (In the past some users set limits on the size of images docker can pull) If there is still only one docker container, let me know and I can help further. Thanks!

Chris Pitchford

unread,
Sep 27, 2016, 2:18:44 PM9/27/16
to Lucida Users
Thanks so much for the generous offer!

I found out that System Integrity Protection in OS X (now macOS) was preventing installations into /usr/bin

Instead of fighting the system, I downloaded VirtualBox and built a Docker server on Ubuntu 14.04 and I'm downloading the source from Github and I get disconnected repeatedly with the following error:

RPC failed: result=56, HTTP code = 200MiB | 652,00 KiB/s

The last part is the only part that changes. I think I need to fix my VB (or Ubuntu rusty) networking!

Can I download just the "tools/deploy" folder for the Dockerfiles? Thanks...

Yunsheng Bai

unread,
Sep 27, 2016, 3:00:28 PM9/27/16
to Lucida Users
You can go to the main page (https://github.com/claritylab/lucida), click the green button "Clone or download", and then click "Download ZIP" which is less than 100MB. Thanks!

Chris Pitchford

unread,
Sep 28, 2016, 12:58:37 PM9/28/16
to Lucida Users
I generated and added my ssh keys as there is a bug in Ubuntu's GIT. 

I was able to provision the VirtualBox server with the Kubecetes pods. The docker pull and wikipedia index download took a while(!) but everything appeared okay.

I started the script `./start_services.sh` and the last pod to be created was mongodb and then the guest system stopped responding. Sending term and shutdown signals failed. I'll restart everything and try again.

I think I didn't have enough RAM set for the guest system. I had 2GB RAM and 200GB disk space initially. I'll up the RAM. If it still fails, I'll be glad to share my setup details and try to help debug. 

Thanks again!

Chris Pitchford

unread,
Sep 28, 2016, 1:20:51 PM9/28/16
to Lucida Users
I increased the RAM to the max I have available: ~5.75GB

The mongo container creation is failing. I ran `sudo kubectl describe pod mongo-controller-fiwt7` and got:

23s 23s 1 {kubelet 127.0.0.1} spec.containers{mongo} WarningFailed Failed to pull image "mongo": failed to register layer: rename /var/lib/docker/image/aufs/layerdb/tmp/layer-288711088 /var/lib/docker/image/aufs/layerdb/sha256/050d573ca3cade2a4d089d8c371fc0cd924be837ea5521f499a1ce5853e6177b: directory not empty


I'll try to manually clear the cache/directory, stop and restart everything. 

::chris

Chris Pitchford

unread,
Sep 28, 2016, 4:29:38 PM9/28/16
to Lucida Users
Restarting Docker after deleting the offending directory was apparently the key to getting MongoDB to install.

I'm not getting any errors from any servers, but when I try to view the UI at http://127.0.0.1:8080/ui/ I get the following error.

"message": "no endpoints available for service \"kubernetes-dashboard\""

I'll start researching that (possibly in another thread), but I welcome suggestions!! Thanks!

::chris

Chris Pitchford

unread,
Sep 28, 2016, 5:10:51 PM9/28/16
to Lucida Users
No errors!

But, now I have to find a version of the tutorial (https://github.com/claritylab/lucida/blob/master/tutorial.pdf) that describes endpoints and how to access the pages as shown in the tutorial.

::chris

Yunsheng Bai

unread,
Sep 29, 2016, 9:50:07 PM9/29/16
to Lucida Users
Glad to hear that. Let me know what I can further help. Thanks!

Lokender Kaushik

unread,
Oct 13, 2016, 4:19:01 AM10/13/16
to Lucida Users
Thanks Chris, Yunsheng - your posts will definitely help newbies like me. I tried to run the set up from mac, but kept running into one problem or other.
I will try virtualbox and update with the results.

Lokender Kaushik

unread,
Oct 18, 2016, 11:06:41 AM10/18/16
to Lucida Users
I finally got it working using virtualbox. My objective was local dev - so guess docker wasn't an option?
Followed instructions given in Read Me. Took around 4-5 hours - but no troubles.

The voice capture is horrible though, guess that is because i am using virtualbox.

Thanks again everyone!

auon kazmi

unread,
Jan 19, 2017, 6:57:50 PM1/19/17
to Lucida Users
Hey Bai,
I created a docker image using Dockerfile present in the lucida/ directory. Now, I just want this to run on my local machine. So, I created a container out of this image. Could you please suggest me what command is analogous to make start_all when using docker container.
P.S. I don't have to use Kubernetes as I am just running it on my own machine.

Thanks!
Auon

André Scholten

unread,
Jan 11, 2018, 5:30:40 PM1/11/18
to Lucida Users
did try to install lucida twice from scratch on a VirtualBox with Ubuntu 16.04 LTS, but keep ending on 
asrmaster-controller-2oj85   0/1       CrashLoopBackOff
ms-controller-ds7xl          0/1       CrashLoopBackOff
we-controller-f76is          0/1       CrashLoopBackOff

I definitely have no clue what to do anymore. Any help?
Reply all
Reply to author
Forward
0 new messages