Re: OIDC role mapping?

2019-12-03 Thread Benjamin Guillon
Hi,

I'd gladly know if that's possible as well.
So far in our tests (keycloak OIDC and OKD 3.11 as well) we did not manage to 
do it.

Best regards,
-- 
Benjamin Guillon

- Mail original -
De: "Jon Stanley" 
À: "users" 
Envoyé: Mardi 3 Décembre 2019 06:20:07
Objet: OIDC role mapping?

Is it possible to map roles based on OpenID claims? I've successfully
got a cluster authenticating with OIDC, but I'm wondering if I can do
authorization over there too :). My IDP that I'm using for testing is
Keycloak, so that should be the easiest thing to do, right? :). I
can't find any documentation or enhancement proposal about that.

___
users mailing list
users@lists.openshift.redhat.com
http://lists.openshift.redhat.com/openshiftmm/listinfo/users

smime.p7s
Description: S/MIME Cryptographic Signature
___
users mailing list
users@lists.openshift.redhat.com
http://lists.openshift.redhat.com/openshiftmm/listinfo/users


CephFS external storage

2019-10-08 Thread Benjamin Guillon
Hey there,

I'm investigating possible alternatives to NFS for ReadWriteMany capable 
external storage.
We're running OKD 3.11 and we are already using Ceph RBD succesfully.

However, since we would like RWM capabilities we were wondering if it was 
possible to use CephFS (which supports RWM and is available in Kubernetes 
Vanilla since v1.5+).
Anyone knows if that's possible or will be possible in the near future? Ideally 
we'd like to be able to create a storage class for dynamic provisioning but 
manual provisioning would be a good start.
Any other viable alternatives suggestions to NFS?

Thanks a lot!
-- 
Benjamin Guillon
CNRS/IN2P3 Computing Center
21 Avenue Pierre de Coubertin, CS70202
69627 Villeurbanne Cedex, France


smime.p7s
Description: S/MIME Cryptographic Signature
___
users mailing list
users@lists.openshift.redhat.com
http://lists.openshift.redhat.com/openshiftmm/listinfo/users


Re: CephFS external storage

2019-10-10 Thread Benjamin Guillon
Hi Samuel, 

Thanks for this information. 
I'll give it a try then! 

Best regards, 
-- 
Benjamin Guillon 
CNRS/IN2P3 Computing Center 
21 Avenue Pierre de Coubertin, CS70202 
69627 Villeurbanne Cedex, France 


De: "Samuel Martín Moro"  
À: "Benjamin Guillon"  
Cc: "users"  
Envoyé: Mardi 8 Octobre 2019 10:30:31 
Objet: Re: CephFS external storage 

Hi, 

CephFS works on OKD 3.11 -- and probably earlier. 
Kubernetes docs are pretty much accurate dealing with CephFS on OpenShift. 
You'ld need a specific provisioner, with admin privileges over your Ceph 
cluster (creating CephFS volumes would create a Ceph keyring with limited 
privileges, such as your Pods may not access shares they're not meant to use) 

One small detail though: I couldn't manage to make it work using the cephfs 
kernel driver. 
If you end up unable to read/write your shares, try installing ceph-fuse 
instead ( [ https://github.com/openshift/origin/issues/21778 | 
https://github.com/openshift/origin/issues/21778 ] ) 


Other alternatives, glusterfs obviously. Viable: I wouldn't go there. 
Stuff like CIFS shares. Though can break some applications (eg: jenkins, low 
posix compliance) 
Haven't had much time to extensively test cephfs. I seem to remember the MDS 
leak (13.x) 
All in all, there's no perfect solution. 

Depending on your application, if you're lucky, another take on it could be to 
look into s3 (radosgw) integration. 


Regards. 

On Tue, Oct 8, 2019 at 9:32 AM Benjamin Guillon < [ 
mailto:benjamin.guil...@cc.in2p3.fr | benjamin.guil...@cc.in2p3.fr ] > wrote: 


Hey there, 

I'm investigating possible alternatives to NFS for ReadWriteMany capable 
external storage. 
We're running OKD 3.11 and we are already using Ceph RBD succesfully. 

However, since we would like RWM capabilities we were wondering if it was 
possible to use CephFS (which supports RWM and is available in Kubernetes 
Vanilla since v1.5+). 
Anyone knows if that's possible or will be possible in the near future? Ideally 
we'd like to be able to create a storage class for dynamic provisioning but 
manual provisioning would be a good start. 
Any other viable alternatives suggestions to NFS? 

Thanks a lot! 
-- 
Benjamin Guillon 
CNRS/IN2P3 Computing Center 
21 Avenue Pierre de Coubertin, CS70202 
69627 Villeurbanne Cedex, France 
___ 
users mailing list 
[ mailto:users@lists.openshift.redhat.com | users@lists.openshift.redhat.com ] 
[ http://lists.openshift.redhat.com/openshiftmm/listinfo/users | 
http://lists.openshift.redhat.com/openshiftmm/listinfo/users ] 





-- 
Samuel Martín Moro 
{EPITECH.} 2011 

"Nobody wants to say how this works. 
Maybe nobody knows ..." 
Xorg.conf(5) 



smime.p7s
Description: S/MIME Cryptographic Signature
___
users mailing list
users@lists.openshift.redhat.com
http://lists.openshift.redhat.com/openshiftmm/listinfo/users


Re: OKD 3.11 Router sharding and exposed routes

2020-05-13 Thread Benjamin Guillon
Hello again,

Additionaly, is there a way to set a default label on route creation?
Read about blueprints somewhere but I didn't find any documentation on how to 
use this.

Thanks a lot !
-- 
Benjamin Guillon
CNRS/IN2P3 Computing Center
21 Avenue Pierre de Coubertin, CS70202
69627 Villeurbanne Cedex, France

- Mail original -
De: "Benjamin Guillon" 
À: "users" 
Envoyé: Mardi 12 Mai 2020 17:20:24
Objet: OKD 3.11 Router sharding and exposed routes

Hello there,

I'm playing with router sharding and I don't know how to "unexpose" a route on 
a router.
Is there a command for this?

I wish to leverage route labels in order to pick which route gets exposed where.
Anyone managed to do this?

The only way I found is deleting the route and creating it again with a new 
label (to expose it to another router for instance).

I'm running a OKD 3.11 cluster.

Thanks for the feedback.
-- 
Benjamin Guillon
___
users mailing list
users@lists.openshift.redhat.com
http://lists.openshift.redhat.com/openshiftmm/listinfo/users

smime.p7s
Description: S/MIME Cryptographic Signature
___
users mailing list
users@lists.openshift.redhat.com
http://lists.openshift.redhat.com/openshiftmm/listinfo/users


OKD 3.11 Router sharding and exposed routes

2020-05-12 Thread Benjamin Guillon
Hello there,

I'm playing with router sharding and I don't know how to "unexpose" a route on 
a router.
Is there a command for this?

I wish to leverage route labels in order to pick which route gets exposed where.
Anyone managed to do this?

The only way I found is deleting the route and creating it again with a new 
label (to expose it to another router for instance).

I'm running a OKD 3.11 cluster.

Thanks for the feedback.
-- 
Benjamin Guillon

smime.p7s
Description: S/MIME Cryptographic Signature
___
users mailing list
users@lists.openshift.redhat.com
http://lists.openshift.redhat.com/openshiftmm/listinfo/users


Slides from Openshift Commons briefing July 25 2019

2020-09-30 Thread Benjamin Guillon
Hello,

Was going back through this video of the Openshift commons briefing:
https://www.openshift.com/blog/introduction-fedora-coreos-fcos-with-benjamin-gilbert-and-ben-breard-red-hat-openshift-commons-briefing

But the slides are not available anymore ...
https://blog.openshift.com/wp-content/uploads/Fedora-CoreOS-OpenShift-Commons-Briefing-July-25-2019.pdf

Would it be possible to get them back online?

Thanks !
-- 
Benjamin Guillon


smime.p7s
Description: S/MIME Cryptographic Signature
___
users mailing list
users@lists.openshift.redhat.com
http://lists.openshift.redhat.com/openshiftmm/listinfo/users


Installing packages on OKD 4 nodes

2020-10-23 Thread Benjamin Guillon
Hello,

I'm deploying an OKD4 cluster on Openstack. 
I wish to configure NTP on my nodes and for that I need to install a PTP 
dependency: linuxptp.
And enable the kvm_ptp module in the kernel.

However, I couldn't manage to install the package through ignition.
I had to do it manually with rpm-ostree: rpm-ostree install linuxptp.

Am I missing something here?
How am I supposed to provide packages or drivers cluster wide through Ignition?
Can such a task be done through the MachineConfig Operator?

Thanks for the help!
-- 
Benjamin Guillon
CNRS/IN2P3 Computing Center
21 Avenue Pierre de Coubertin, CS70202
69627 Villeurbanne Cedex, France


smime.p7s
Description: S/MIME Cryptographic Signature
___
users mailing list
users@lists.openshift.redhat.com
http://lists.openshift.redhat.com/openshiftmm/listinfo/users


Re: Installing packages on OKD 4 nodes

2020-10-28 Thread Benjamin Guillon
Hi Joel, 

Well, no: /dev/ptp0 does not magically exists. Wouldn't that be too easy? :) 

As for a FCOS specific documentation regarding NTP/PTP I'm afraid I didn't find 
any. 

Best, 
-- 
Benjamin 

De: "Joel Pearson"  
À: "Benjamin Guillon"  
Cc: "users"  
Envoyé: Mercredi 28 Octobre 2020 13:56:05 
Objet: Re: Installing packages on OKD 4 nodes 

Hi Benjamin, 
Those docs you’ve mentioned are for regular fedora not fedora coreos I believe 
which I’m pretty sure are very different. 

So I presume you have checked that /dev/ptp0 doesn’t already magically exist? 

Thanks, 

Joel 

Sent from my iPhone 



On 28 Oct 2020, at 11:45 pm, Benjamin Guillon  
wrote: 





BQ_BEGIN

Hi Joel, 

Thanks for the reply :) 

I did give a try to the Openshift PTP operator and works well aside from the 
fact that I can't use it here since I'm not running on Baremetal. 

Our cluster indeed runs on our in-house Openstack platform. 

Usually we use the KVM PTP module with something like: 
refclock PHC /dev/ptp0 poll 2 
In the chrony.conf file. 
But that's for our usual CentOs based VMs, not FCOS :/ 

So now I'm just trying to reproduce that setup on FCOS. 
I found this Fedora documentation earlier about PTP 
https://docs.fedoraproject.org/en-US/fedora/rawhide/system-administrators-guide/servers/Configuring_PTP_Using_ptp4l/
 
Where they mention this linuxptp package, hence my questions. 

If I can't manage this, I'll resort to using standard NTP instead of PTP. 

Best, 
Benjamin 

De: "Joel Pearson"  
À: "Benjamin Guillon"  
Cc: "users"  
Envoyé: Mercredi 28 Octobre 2020 12:56:19 
Objet: Re: Installing packages on OKD 4 nodes 

Ahh I found the support article that talks about OpenShift 4 and PTP 

[ https://access.redhat.com/solutions/5106141 | 
https://access.redhat.com/solutions/5106141 ] 

If you don't have access to that solution the crux of it is that the PTP 
operator is for baremetal nodes (so probably not you, as you mentioned 
OpenStack). 

The chrony config they mention is: 

$ cat << EOF | base64 -w0
refclock PHC /dev/ptp0 poll 3 dpoll -2 offset 0
driftfile /var/lib/chrony/drift
makestep 1.0 3
rtcsync
logdir /var/log/chrony
EOF 

On Wed, 28 Oct 2020 at 22:32, Joel Pearson < [ 
mailto:japear...@agiledigital.com.au | japear...@agiledigital.com.au ] > wrote: 

BQ_BEGIN

Hi Benjamin, 
Have you checked if you actually need it? At least enterprise openshift 4.x 
already had ptp support in the kernel (without a module), as I bumped into it 
earlier in the year for PTP Azure syncing, I opened a support ticket and it 
turned out I just needed this in chrony.conf 

refclock PHC /dev/ptp0 poll 3 dpoll -2 offset 0 
So I think it'd be worth checking if you already have /dev/ptp0 available 
before installing linuxptp. I realise OKD uses Fedora Core OS instead of RedHat 
Core OS, so the default kernel modules might be different. 

Here are some docs for [ 
https://docs.okd.io/latest/installing/install_config/installing-customizing.html#installation-special-config-crony_installing-customizing
 | configuring chrony ] , I think you just need to switch the iburst line for 
the refclock one. 

Otherwise, if the PTP support you need is more complicated than I needed on 
Azure, you could potentially look at the specific [ 
https://docs.okd.io/latest/networking/multiple_networks/configuring-ptp.html | 
PTP operator ] in the OKD docs. 

Hope this helps. 

Thanks, 

Joel 


On Sat, 24 Oct 2020 at 03:00, Benjamin Guillon < [ 
mailto:benjamin.guil...@cc.in2p3.fr | benjamin.guil...@cc.in2p3.fr ] > wrote: 

BQ_BEGIN
Hello, 

I'm deploying an OKD4 cluster on Openstack. 
I wish to configure NTP on my nodes and for that I need to install a PTP 
dependency: linuxptp. 
And enable the kvm_ptp module in the kernel. 

However, I couldn't manage to install the package through ignition. 
I had to do it manually with rpm-ostree: rpm-ostree install linuxptp. 

Am I missing something here? 
How am I supposed to provide packages or drivers cluster wide through Ignition? 
Can such a task be done through the MachineConfig Operator? 

Thanks for the help! 
-- 
Benjamin Guillon 
CNRS/IN2P3 Computing Center 
21 Avenue Pierre de Coubertin, CS70202 
69627 Villeurbanne Cedex, France 
___ 
users mailing list 
[ mailto:users@lists.openshift.redhat.com | users@lists.openshift.redhat.com ] 
[ http://lists.openshift.redhat.com/openshiftmm/listinfo/users | 
http://lists.openshift.redhat.com/openshiftmm/listinfo/users ] 

BQ_END



BQ_END



BQ_END




smime.p7s
Description: S/MIME Cryptographic Signature
___
users mailing list
users@lists.openshift.redhat.com
http://lists.openshift.redhat.com/openshiftmm/listinfo/users


Re: Installing packages on OKD 4 nodes

2020-10-28 Thread Benjamin Guillon
Hi Joel, 

Thanks for the reply :) 

I did give a try to the Openshift PTP operator and works well aside from the 
fact that I can't use it here since I'm not running on Baremetal. 

Our cluster indeed runs on our in-house Openstack platform. 

Usually we use the KVM PTP module with something like: 
refclock PHC /dev/ptp0 poll 2 
In the chrony.conf file. 
But that's for our usual CentOs based VMs, not FCOS :/ 

So now I'm just trying to reproduce that setup on FCOS. 
I found this Fedora documentation earlier about PTP 
https://docs.fedoraproject.org/en-US/fedora/rawhide/system-administrators-guide/servers/Configuring_PTP_Using_ptp4l/
 
Where they mention this linuxptp package, hence my questions. 

If I can't manage this, I'll resort to using standard NTP instead of PTP. 

Best, 
Benjamin 

De: "Joel Pearson"  
À: "Benjamin Guillon"  
Cc: "users"  
Envoyé: Mercredi 28 Octobre 2020 12:56:19 
Objet: Re: Installing packages on OKD 4 nodes 

Ahh I found the support article that talks about OpenShift 4 and PTP 

[ https://access.redhat.com/solutions/5106141 | 
https://access.redhat.com/solutions/5106141 ] 

If you don't have access to that solution the crux of it is that the PTP 
operator is for baremetal nodes (so probably not you, as you mentioned 
OpenStack). 

The chrony config they mention is: 

$ cat << EOF | base64 -w0
refclock PHC /dev/ptp0 poll 3 dpoll -2 offset 0
driftfile /var/lib/chrony/drift
makestep 1.0 3
rtcsync
logdir /var/log/chrony
EOF 

On Wed, 28 Oct 2020 at 22:32, Joel Pearson < [ 
mailto:japear...@agiledigital.com.au | japear...@agiledigital.com.au ] > wrote: 



Hi Benjamin, 
Have you checked if you actually need it? At least enterprise openshift 4.x 
already had ptp support in the kernel (without a module), as I bumped into it 
earlier in the year for PTP Azure syncing, I opened a support ticket and it 
turned out I just needed this in chrony.conf 

refclock PHC /dev/ptp0 poll 3 dpoll -2 offset 0 
So I think it'd be worth checking if you already have /dev/ptp0 available 
before installing linuxptp. I realise OKD uses Fedora Core OS instead of RedHat 
Core OS, so the default kernel modules might be different. 

Here are some docs for [ 
https://docs.okd.io/latest/installing/install_config/installing-customizing.html#installation-special-config-crony_installing-customizing
 | configuring chrony ] , I think you just need to switch the iburst line for 
the refclock one. 

Otherwise, if the PTP support you need is more complicated than I needed on 
Azure, you could potentially look at the specific [ 
https://docs.okd.io/latest/networking/multiple_networks/configuring-ptp.html | 
PTP operator ] in the OKD docs. 

Hope this helps. 

Thanks, 

Joel 


On Sat, 24 Oct 2020 at 03:00, Benjamin Guillon < [ 
mailto:benjamin.guil...@cc.in2p3.fr | benjamin.guil...@cc.in2p3.fr ] > wrote: 

BQ_BEGIN
Hello, 

I'm deploying an OKD4 cluster on Openstack. 
I wish to configure NTP on my nodes and for that I need to install a PTP 
dependency: linuxptp. 
And enable the kvm_ptp module in the kernel. 

However, I couldn't manage to install the package through ignition. 
I had to do it manually with rpm-ostree: rpm-ostree install linuxptp. 

Am I missing something here? 
How am I supposed to provide packages or drivers cluster wide through Ignition? 
Can such a task be done through the MachineConfig Operator? 

Thanks for the help! 
-- 
Benjamin Guillon 
CNRS/IN2P3 Computing Center 
21 Avenue Pierre de Coubertin, CS70202 
69627 Villeurbanne Cedex, France 
___ 
users mailing list 
[ mailto:users@lists.openshift.redhat.com | users@lists.openshift.redhat.com ] 
[ http://lists.openshift.redhat.com/openshiftmm/listinfo/users | 
http://lists.openshift.redhat.com/openshiftmm/listinfo/users ] 





BQ_END





smime.p7s
Description: S/MIME Cryptographic Signature
___
users mailing list
users@lists.openshift.redhat.com
http://lists.openshift.redhat.com/openshiftmm/listinfo/users