@otmarjr Regarding the server installation, what error do you get when you run rasactl start avcsti --values-file ~/values.ym can you share the log with me?
The process after you run this command should take a few minutes. Another important thing to do if you are installing Rasa X on a GCP or AWS server is updating your firewall rules in your instance configuration and opening up the port that Rasa X is running on.
Last time I tried on my company’s server (on premises, not cloud-based), I got the following error:
Error while waiting for Rasa X Database migration status, error: context deadline exceeded
Could you please guide me on how to get the installation log? I’m aware only of the rasactl logs command, which then asks me to select a pod to get the logs from.
I’ve succesfully installed it by following the directions and links available at Rasa X 1.0.x’ page: Docker Compose Installation
Regarding marking a thread solution, I ask one clarification: REI installer seems to still be very problematic and not working properly, what makes me think there’s no current stable solution for it. Should’nt there be a solution only when REI issues are finally resolved? It’s important to note that the solution I’ve used in the past is currently deprecated for Rasa X 1.1
I agree with @otmarjr and raven99p
I’m still using old method in production (docker-composer with pip install) because i have problems with REI/Helm installation.
REI is a great idea but in sometimes, i just need a simple installation without kubernet or something like that.
Sometimes, i help people without technical expertise to install Rasa and Rasa X in their computers. Pip method installation is perfect, its simple and work fine
I like the news methods but i disagreed to deprecated Pip install method
I should also like to take this chance to mention that i thing the Rasa X Matrix Doc is wrong
@otmarjr is your current installation as mentioned in the shared link working fine? I guess you are aware that there is no known Docker-compose installation. Good Luck!
It worked for me to install Rasa X 1.0.x using Docker compose following the instructions provided by the link I posted above. However, we know this method is now deprecated for Rasa X 1.1.x onwards.
So my fellow Rasa X users and myself are all waiting for a stable REI release. Hopefully, this thread will have a solution soon, when Rasa Team fix REI. If you know anyway I could help or contribute for this fix, let me know! I’d love to spend my week off contributing to this greater goal!
To be honest, since I opened this thread, I’ve got my way by going with Rasa X 1.0.x and Docker compose.
Based on itsjhonny reply. Rasa 3.x does not work with Rasa X 1.1.x. In that situation, I’d stick with Rasa X 1.0.x (using Docker compose installation method) until there’s a proven way to install a Rasa X version compatible with Rasa 3.x. Unless you’re 100% sure there’s something you really need and is only available on Rasa X 1.1.x, my best guess is that Rasa X 1.0.x is the way to go so far.
I guess you, just like me, want to use the most up to date version, due to their new features and cool stuff. However, experience shows that sometimes we must take a step back, focus on our chatbot using stable versions and,d after spending some time with our chatbot and end users, come back and see if there’s some something new for our toolchain. That’s the strategy I’ve succesfully deployed so far
REI is a bootstrapper to install a local K8S Cluster and its requirements for the rasactl to work so that a deployment of RASA OSS / X is easy possible and thats it.
if there are problems with your local REI provisioned KIND setup its good to look into logs of docker on OS level to see what problems are there (disk full, clashing docker installations, previous remnants of interfering deployments on SDN of docker)
as ultimate solution do a reinstallation of docker to clear up previous installations of other things that might have been installed.
rasactl uses the REI’s provisioned local K8S Cluster to deploy you local installation.
for more details on rasactl on how to debug / work with local mounted volumes / training models
last but not least
dont use rasactl / REI for production installations
default production deployment are via helm to k8s.
if you are up to more tinkering and want it really really fast quick and dirty.
pip is the way - use pyhton virtual envs to keep your sanity.
Hey @RASADSA . Could you explain why we shouldn’t use rasactl for production install?
(I want to share my assistant using RASA X on GCP and rasactl installs the version, which seems not to work properly with model, trained on RASA 3.0)