Forum Replies Created
-
Posts
-
1. You have sudo on all VMs. Please see example notebooks, many of them show how to install additional packages via yum or dnf, depending on the distribution. E.g. look at the GPU notebook.
2. You may find the following article useful to better understand distinctions between services. You may also find this article useful to understand the differences between interfaces. In general FABlib does the right thing most of the time if you just tell it you want L2 or L3 and you don’t need to worry about L2STS, L2PTP or L2Bridge. There are some corner cases, but it doesn’t sound like you are near those.
3, Which one (L2 or L3) you pick depends entirely on what you want to do. L2 services require you to set up your own L3. L3 (IPv4 and IPv6) services come with prepackaged routing that is done internal to FABRIC dataplane outside of the VM. You don’t need to be on different sites, you can be in the same site.
As I mentioned you have full sudo rights. However be cognizant of not removing the default route that usually points to the management interface – you will cut off your access. This applies regardless of whether you build your own L3 or use one of our FABNet services.
Also be cognizant of which container type you are starting – we now offer several. For experimenters the ‘bleeding edge’ is worth trying as it should have all the latest released versions
No, as per documentation, Project Lead role is needed to create new projects (I strongly suggest looking over that article, particularly its last section). If the project has been created, they can be members or, if they want to be able to add other members like their students, they can be project owners. All members of the projects (owners, regular members) have the same rights with respect to creating slices. Sliver keys should be generated on a per person basis.
If they wish to access each others’ slices, they can add each others’ public sliver SSH keys into VMs in their slices (this can be easily done using ‘node.execute(command)’) after the slice is created. The keys need to be appended to ~/.ssh/authorized_keys file in each VM. We are adding a more programmatic way to do it, but it is not currently available.
Understood,
Please encourage them also to ask questions here on the forums. Suggestions for features and documentation are always welcome (we may not do it right away but we do take notes).
Looks to me like you are all set – your fabric_rc looks as expected.
I am not sure the issue with loading the portal/slice builder. The usual advice is to log out, remove all cookies that start with ‘fabric’, then log back in.
Giorgio,
We have been observing some instability in the bastion hosts – thank you for bringing it to our attention (you happen to have run into it into one of them but it appears both are affected). We think we know the issue and will update their configuration to avoid problems of this kind.
Acheme,
We did a first-order assessment – we should be able to do this (probably) – however the reports of OpenStack support for this (which we need in order to enable for experiments) suggest there may be bugs. We will need to test and make sure it doesn’t affect the stability of the system. We will also need to assess what actions the control software needs to take to support it. Thank you for bringing this to our attention – we will investigate further and report on this thread.
Fengping,
We are a bit thinly staffed this week, but I will ask someone on operations team to look into it. Can you post any relevant configuration bits – what you may have done right before you lost access, post boot scripts or node.execute() scripts? The most likely cause is a change in routing inside the node that fried the default route pointing back to the management interface. Other options are possible, of course.
Hello,
A couple of things:
1. Probably not the best strategy to create a fabric ssh config that includes your sliver address – this address will change from slice to slice and you don’t want to be editing the config file every time, but this may be a personal preference.
2. Jupyter Hub is ‘outside’ of FABRIC just like your laptop – it is hosted in Google Cloud
3. FABRIC has multiple bastion hosts (currently two with the two addresses you noted) which round-robin resolve to the name bastion.fabric-testbed.net. One of them apparently is giving you trouble – not clear why – I can see your key is properly installed there. We will look into it, thank you for reporting.
