Hello, I’ve a very dumb question about the difference between eos server and eos client.
I’m starting a test case using only one node first to learn lessons and then plan my environment, but my test machine is a Ubuntu 22.04 and seems that the server must be a CentOS.
I understand that I need a Server+Client on this machine and than I can have only the client side on other machines, is that correct?
We build the eos client which includes the eos client console and the fuse daemon also for the Ubuntu distribution but we don’t do this for the eos-server package which includes (MGM, FST, MQ daemons) and neither for the eos-quakrdb. The reason for this is that all our infrastructure is RHEL based and there are quite a few dependencies that also need to be build for the server part to work properly.
Therefore, my suggestion would be to use docker containers with CentOS7 as a base image and experiment inside a container rather than on the host machine which from what I understand is Ubuntu. The performance penalty is negligible for the vast majority of real world use-cases.
You can use an already prepared image if you have access to the CERN Gitlab Registry:
docker pull gitlab-registry.cern.ch/dss/eos/eos-ci:5.1.29
Hope this helps!
Thanks for the explanation. I got it.
Since I have 2 SSD’s and 2 HDD’s on my machine, I built a dual boot on the SSD’s, one with the original Ubuntu 22.04 and a new one with CentOS-9-Stream and used one HDD for each system.
Tested and it’s working fine!
Now I’ll proceed with my EOS Server (MGM, MQ and FST) and as soon as I have this working, I’ll install the client only in another machine.