[Debnil Sur]

That gives me this error:

`
—————————————————————————
TypeError Traceback (most recent call last)
Cell In[6], line 54
51 prompt_to_destroy_all_slices()
53 fablib.show_config()
—> 54 output = fablib.list_sites(force_refresh=True)
56 None

TypeError: list_sites() got an unexpected keyword argument ‘force_refresh’

<code></code>`

I’m using the Fall 2023 server (the default).

[Debnil Sur]

To provide a reproducible example:

from fabrictestbed_extensions.fablib.fablib import FablibManager as fablib_manager
fablib = fablib_manager()
slice_name = 'ei-2node'
slice = fablib.get_slice(name=slice_name)
node_name = 'ei11'
node = slice.get_node(name=node_name)
node.execute(f’sudo apt update’)

resulting in:

sudo: unable to resolve host e0a0baa2-fbe3-4d6b-92b9-6d1c74b93897-ei11: Name or service not known

[Debnil Sur]

Is there a reason the behavior would differ between SSH’ing onto the node and executing commands through the Jupyter notebook? It’s not letting me upload a screenshot, but the behavior is roughly:

node.execute(f’sudo apt update’)

sudo: unable to resolve host e0a0baa2-fbe3-4d6b-92b9-6d1c74b93897-ei11: Name or service not known

[Debnil Sur]

I have two VMs on DALL. Their management IP addresses are 2001:400:a100:3000:f816:3eff:fed0:4779 and 2001:400:a100:3000:f816:3eff:fe62:fc9a.

[Debnil Sur]

Sadly, I’m getting the same sudo error in other FABRIC sites (not NCSA). I’d like to use Ubuntu; I’m trying to benchmark an existing project, built on Linux and which requires cloning from Github. I’d imagine I could run it on Rocky Linux, but  I don’t want to have to write additional installation scripts for various dependencies. Is there any workaround, or is Rocky my best bet?

[Debnil Sur]

I’m going in circles on this workaround. To change the DNS server, I thought I would need to edit /etc/resolve.conf. This requires root privileges. That isn’t possible if sudo doesn’t work. What’s the way around this?

[Author]

Posts