Provide Fedora setup instructions

[ncoghlan]

The Vagrant setup instructions unfortunately don't work on Fedora: http://fabric8.io/guide/getStarted/vagrant.html

Specifically, the local DNS setup doesn't work, which means the reader is left completely on their own figuring out what to do about it.

[gashcrumb]

I have it working here via a global dnsmasq instance, let me take a quick gander though at a simpler setup, will document it here and then we can get that rolled into the docs...

[gashcrumb]

Okay, here's the raw steps that should work for a typical Fedora installation where network connections are managed by network manager, BTW you can do this even after the vagrant image is running:

1. Edit /etc/NetworkManager/NetworkManager.conf and enable DNS caching under the [main] section, for example:

[main]
plugins=ifcfg-rh,ibft
dns=dnsmasq

2. Add the landrush config to NetworkManager's dnsmasq configuration under /etc/NetworkManager/dnsmasq.d in a file called vagrant-landrush.conf:

server=/vagrant.f8/127.0.0.1#10053

3. Restart NetworkManager via systemctl restart NetworkManager

Doing a dig fabric8.vagrant.f8 should give you the proper answer at this point:

] dig fabric8.vagrant.f8

; <<>> DiG 9.10.2-P4-RedHat-9.10.2-5.P4.fc22 <<>> fabric8.vagrant.f8
;; global options: +cmd
;; Got answer:
;; ->>HEADER<<- opcode: QUERY, status: NOERROR, id: 40548
;; flags: qr aa rd ra; QUERY: 1, ANSWER: 1, AUTHORITY: 0, ADDITIONAL: 0

;; QUESTION SECTION:
;fabric8.vagrant.f8.        IN  A

;; ANSWER SECTION:
fabric8.vagrant.f8. 0   IN  A   172.28.128.4

;; Query time: 1 msec
;; SERVER: 127.0.0.1#53(127.0.0.1)
;; WHEN: Tue Oct 20 10:48:49 EDT 2015
;; MSG SIZE  rcvd: 52

[gashcrumb]

@ncoghlan let me know if that works for you and I'll get the docs updated, thanks!

[ncoghlan]

Unfortunately not. It did change /etc/resolv.conf to use the local resolver, but it only changed the dig error message from NXDOMAIN to the following:

$ dig fabric8.vagrant.f8

; <<>> DiG 9.10.3-RedHat-9.10.3-1.fc23 <<>> fabric8.vagrant.f8
;; global options: +cmd
;; connection timed out; no servers could be reached

Normal DNS lookups are working fine, it's only the local lookup that is having problems. The landrush setting looks like it should be right:

$ cat /etc/NetworkManager/dnsmasq.d/vagrant-landrush 
server=/vagrant.f8/127.0.0.1#10053

[ncoghlan]

There doesn't appear to be anything listening on port 10053, so it may be a problem with landrush not running correctly.

[gashcrumb]

Ah yeah, that'd definitely be a problem... You should see landrush initialize as part of vagrant up, be worth validating that there's no obvious errors from that in the output.

[ncoghlan]

I'm trying again, having explicit started vagrant landrush start before running vagrant up.

SELinux is in permissive mode, so if it still doesn't work, then the remaining likely culprits would be:

• a compatibility issue with the vagrant-libvirt backend
• a compatibility issue with running vagrant as root (since the libvirt vagrant backend uses the system virt session)

[gashcrumb]

Gotcha. I'm still running with the VirtualBox backend here since originally the vagrant image didn't support libvirt.

[ncoghlan]

Either setting up the local resolver or starting landrush explicitly appears to have made things worse - vagrant up is now getting stuck at "Waiting for domain to get an IP address...".

I'll keep digging tomorrow.

[ncoghlan]

Retrying again today, the landrush daemon again didn't start, and there wasn't any output at all relating to landrush in the vagrant up output.

Further investigation found vagrant-landrush/landrush#20, so it looks like anything based on landrush simply won't work with the libvirt backend for the time being.

@strzibny has a PR adding vagrant-libvirt support at vagrant-landrush/landrush#124

[ncoghlan]

I'm trying to figure out from the OpenShift console (which has its IP address displayed during vagrant up) what the IP address of fabric8.vagrant.f8 would be, so I can add it to dnsmasq manually, but none of the addresses I can find will resolve from the host.

[jimmidyson]

It should be the public IP of the VM - if you're using the vagrant setup in fabric8 that would be 172.28.128.4.

[ncoghlan]

That's coming back as "503 service unavailable", so I'm guessing the failure of the DNS resolution is meaning some latter aspect of the Fabric8 setup isn't executing properly.

I can't see any obvious errors in the vagrant up output or the OpenShift console, though.

[jimmidyson]

What does oc get routes show? The host field in there needs to match fabric8.vagrant.f8 for the Vagrant box - that should be handled on vagrant up.

[ncoghlan]

@jimmidyson explained on IRC that the OpenShift router needs the hostname in order to route the request to the correct service, so without the landrush/libvirt integration, it's necessary to add the appropriate entry to /etc/hosts.

[ncoghlan]

I think that means the near term Fedora instructions are going to need to be something like "Edit /etc/hosts and add a line like:

172.28.128.4 fabric8.vagrant.f8 vagrant.f8

With any other services also added manually to that line.

[jimmidyson]

Only for fedora & libvirt - landrush works fine with virtualbox afaik.

[ncoghlan]

That's a good point, so a better near term option might be to recommend using VirtualBox even on Fedora, and point out the manual workaround for folks using the native libvirt backend.

[strzibny]

@ncoghlan so whats missing in landrush if my patch got merged? can I be of any help here?

[jimmidyson]

@strzibny Integration with the libvirt dnsmasq server would be great.

[oscerd]

I have similar problem on Centos 7. I have to add addresses manually in /etc/hosts