You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
As an operator I want to clearly define a minimum set of requirements to be met before a node can become part of SCIONLab as infrastructure node.
State
The current version of scion-tutorials/content/join_infrastructure.md is highly outdated mainly (but not only) in parts about operating system and firewall rules.
[A] There also statements which we don't want to have, like you can start as small as dedicating only a simple commodity PC. In this case user should be running an user AS as this is highly probable a commodity PC will be an old&cheap machine under someone's desk without any operational support at all.
[B] Currently we require minimum of 4 GB of RAM. This is not enough in many cases and causes us to run at the edge what also increases operational costs in case strange things happen. Nowadays a reasonable requirement should be 8 GB of ram and at least 4 vCPUs.
[C] A model operating system for the infrastructure node is currently Ubuntu 18.04.
[D] border router node(s) must have a public static IP should be extended to a requirement where all the nodes have static IPs. Otherwise that would mean for intra-AS communication we need to deal with dynamic IPs what sounds crazy.
[E] The sentence it may be enough if a firewall allows return traffic to the same port sounds very vague. For people dealing with firewall configurations it's more than enough to be told something like "incoming traffic from X to Y over port Z"
[F] Address ranges for management SSH traffic are very incorrect.
[G] The last option in we can also operate the connections over a tunnel, e.g. OpenVPN, Wireguard or SSH tunnels should be removed. I cannot imagine operating anything more than "debug & very temporary test" via SSH tunnel as this thing is not properly deployable, monitorable nor operatable.
Further improvements
As a further enhancement I want to provide a "hardened configuration" in case there is a business requirement to run in a harsh/strict environment, given the value is higher than investment needed.
The text was updated successfully, but these errors were encountered:
Background
As an operator I want to clearly define a minimum set of requirements to be met before a node can become part of SCIONLab as infrastructure node.
State
The current version of
scion-tutorials/content/join_infrastructure.md
is highly outdated mainly (but not only) in parts about operating system and firewall rules.[A] There also statements which we don't want to have, like
you can start as small as dedicating only a simple commodity PC
. In this case user should be running an user AS as this is highly probable a commodity PC will be an old&cheap machine under someone's desk without any operational support at all.[B] Currently we require
minimum of 4 GB of RAM
. This is not enough in many cases and causes us to run at the edge what also increases operational costs in case strange things happen. Nowadays a reasonable requirement should be 8 GB of ram and at least 4 vCPUs.[C] A model operating system for the infrastructure node is currently Ubuntu 18.04.
[D]
border router node(s) must have a public static IP
should be extended to a requirement where all the nodes have static IPs. Otherwise that would mean for intra-AS communication we need to deal with dynamic IPs what sounds crazy.[E] The sentence
it may be enough if a firewall allows return traffic to the same port
sounds very vague. For people dealing with firewall configurations it's more than enough to be told something like "incoming traffic from X to Y over port Z"[F] Address ranges for management SSH traffic are very incorrect.
[G] The last option in
we can also operate the connections over a tunnel, e.g. OpenVPN, Wireguard or SSH tunnels
should be removed. I cannot imagine operating anything more than "debug & very temporary test" via SSH tunnel as this thing is not properly deployable, monitorable nor operatable.Further improvements
As a further enhancement I want to provide a "hardened configuration" in case there is a business requirement to run in a harsh/strict environment, given the value is higher than investment needed.
The text was updated successfully, but these errors were encountered: