Re: NFS on RDMA
Hean, sound like a cool feature (a bit late in my reply, sorry) I think as far as cloudstack is concerned the feature would be adding nfs options (shouldn't be rdma specific) I don't think we support that right now. On Thu, Dec 24, 2020 at 2:30 AM Hean Seng wrote: > HI > > Is there any way to mount NFS over RDMA . > > We need to do : > > *mount -o proto=rdma,port=20049 * > > for Moiuting NFS over RDMA > > -- > Regards, > Hean Seng > -- Daan
SystemVMs default gateway
Hello, Are SystemVM's meant to have a default gateway? When I SSH directly into the SSVM, I can see it is attached to 3 networks (management, guest, loopback) however there is no gateway set so they are unable to ping externally to the outside world. Was having trouble uploading templates and wondered if this will be the cause. Thank you Pieter
Re: Kubernetes 1.19+
Hi Cristian, I tried to deploy a k8s cluster with k8s version 1.19.7 but it seems to be failing, probably while trying to deploy networking for the cluster. I'll try to debug the issue and will update. Regards, Abhishek From: cristian.c@istream.today Sent: 25 January 2021 13:31 To: users@cloudstack.apache.org Subject: Kubernetes 1.19+ Hello, Did anyone test Kubernetes 1.19 on cloudstack 4.15? Regards, Cristian abhishek.ku...@shapeblue.com www.shapeblue.com 3 London Bridge Street, 3rd floor, News Building, London SE1 9SGUK @shapeblue
Re: CS Kubernetes & persistent storage
sorry for the late answer Niclas, We don't have a solution in ACS for him right now. You probably want to share access to a mount on some persistent storage across VMs so a k8s container can move between VMs and still work on the same data. I don't think a zone wide storage or any other means in ACS would help with this. As you mention, you'll will have to run a nfs server and mount it on all node VMs manually. nice feature request (once you figure it out) On Thu, Dec 17, 2020 at 9:00 AM Niclas Lindblom wrote: > Hi all, > > I am testing the Kubernetes plugin for CS 4.14 and I am trying to figure > out how to manage persistent storage across multiple nodes so a container > can survive being moved from one node to another. The only thing I can > think of to make this work would be a separate NFS server that containers > can mount to, but perhaps there is a better option ? Are there any best > practices around how to implement this in Cloudstack ? > > Regards > > Niclas -- Daan
Re: SystemVMs default gateway
Okay, so I figured out why my system VM's are unable to route traffic. Our network is pretty basic in CloudStack. We have a single shared guest network VLAN2, when I bring up the zone, the system VM's get an IP address from the defaultGuestNetwork which is a routable network on our core router. The System VM can ping the gateway however it cannot reach anything such as internet addresses or other networks shown by the router. Checked the host for IP tables rules and found the below -A FORWARD -o cloudbr0 -m physdev --physdev-is-bridged -j BF-cloudbr0 -A FORWARD -i cloudbr0 -m physdev --physdev-is-bridged -j BF-cloudbr0 -A FORWARD -o cloudbr0 -j DROP -A FORWARD -i cloudbr0 -j DROP -A FORWARD -o brbond0-2 -m physdev --physdev-is-bridged -j BF-brbond0-2 -A FORWARD -i brbond0-2 -m physdev --physdev-is-bridged -j BF-brbond0-2 -A FORWARD -o brbond0-2 -j DROP -A FORWARD -i brbond0-2 -j DROP After removing drop rules the System VM's were able to get traffic flow again. It can download ISO's and Templates but only if I remove those DROP rules. Is this an issue with execution order on Ubuntu 20.04? To allow traffic to the System VM's and then to promptly drop traffic again? Or am I looking at this all the wrong way. An example of our network: management: 10.0.0.0/24 (VLAN1) guest: 10.0.1.0/24 (VLAN2) Followed all the CloudStack documentation using Advanced Networking with Security Groups Thank you Pieter On 10 February 2021 at 11:27, Pieter Koorts wrote: Hello, Are SystemVM's meant to have a default gateway? When I SSH directly into the SSVM, I can see it is attached to 3 networks (management, guest, loopback) however there is no gateway set so they are unable to ping externally to the outside world. Was having trouble uploading templates and wondered if this will be the cause. Thank you Pieter
Re: CS Kubernetes & persistent storage
Thanks Daan, I ended up using Rook Ceph file system across a 3 node cluster with a virtual disk attached to each vm. Works pretty well so far and I can recommend anyone with the same requirement to check it out. Regards Niclas > On 10 Feb 2021, at 22:59, Daan Hoogland wrote: > > sorry for the late answer Niclas, > We don't have a solution in ACS for him right now. You probably want to > share access to a mount on some persistent storage across VMs so a k8s > container can move between VMs and still work on the same data. I don't > think a zone wide storage or any other means in ACS would help with this. > As you mention, you'll will have to run a nfs server and mount it on all > node VMs manually. nice feature request (once you figure it out) > >> On Thu, Dec 17, 2020 at 9:00 AM Niclas Lindblom >> wrote: >> >> Hi all, >> >> I am testing the Kubernetes plugin for CS 4.14 and I am trying to figure >> out how to manage persistent storage across multiple nodes so a container >> can survive being moved from one node to another. The only thing I can >> think of to make this work would be a separate NFS server that containers >> can mount to, but perhaps there is a better option ? Are there any best >> practices around how to implement this in Cloudstack ? >> >> Regards >> >> Niclas > > > > -- > Daan
