If the same Falcon topology is run twice on the same machine, ZFS and data link names will collide. This is extremely unfortunate for shared lab environments.
A workaround is that users can give topologies unique names. There are a few ways to go about sorting this out.
Add a random string or a counter of some kind to topology names.
If we choose the random string, it would need to be somewhat short, as we can easily bump up against OS limits for link names. If we choose the counter, system-wide coordination would be needed.
Use random IDs for link names and ZFS names.
If we do this, features would need to be built into the Falcon CLI utilities to map the random names to topology-relevant names to aid in debugging and general observability.
elaine-oxide
commented
1 month ago
If the same Falcon topology is run twice on the same machine, ZFS and data link names will collide. This is extremely unfortunate for shared lab environments.
A workaround is that users can give topologies unique names. There are a few ways to go about sorting this out.
If we choose the random string, it would need to be somewhat short, as we can easily bump up against OS limits for link names. If we choose the counter, system-wide coordination would be needed.
If we do this, features would need to be built into the Falcon CLI utilities to map the random names to topology-relevant names to aid in debugging and general observability.