The NIST team, @tkphd and @reid-a, did another "road-test" of this lesson.
The learners had a mix of experience at the command line, but all were novices to HPC.
The example cluster requires SSH keys to access, and does not permit password access. SSH key management again consumed a large share of the first part of the lesson. Some learners were idle for a long time while others were getting their keys sorted out, and a few learners declined to complete the connection process. This is a hard issue, connectivity is one of the important novel points for the lesson, so relegating it to the set-up seems wrong, but it tends to serialize, and so consume a lot of time.
Engaging with users who had questions opened up some rabbit holes, which the instructors were perhaps a bit too eager to go down, specifically involving the (abbreviated!) discussion of bash prompts. Possibly requiring Shell novice will help with this.
Our system has a nontrivial layout of the file system, some learners felt this deserved more attention.
We again struggled with focus and scope in the pi.py example, the move (already underway) to replace this with a black-box Amdahl's law demo tool is welcome.
The jargon-busting presentation was well-received.
For instructors sharing the terminal, there was a persistent problem with important commands and context scrolling off the screen when info commands were run. One learner pointed out that the Code Refinery people have a command history tool which helps with this.
The NIST team, @tkphd and @reid-a, did another "road-test" of this lesson.
The learners had a mix of experience at the command line, but all were novices to HPC.
The example cluster requires SSH keys to access, and does not permit password access. SSH key management again consumed a large share of the first part of the lesson. Some learners were idle for a long time while others were getting their keys sorted out, and a few learners declined to complete the connection process. This is a hard issue, connectivity is one of the important novel points for the lesson, so relegating it to the set-up seems wrong, but it tends to serialize, and so consume a lot of time.
Engaging with users who had questions opened up some rabbit holes, which the instructors were perhaps a bit too eager to go down, specifically involving the (abbreviated!) discussion of bash prompts. Possibly requiring Shell novice will help with this.
Our system has a nontrivial layout of the file system, some learners felt this deserved more attention.
We again struggled with focus and scope in the pi.py example, the move (already underway) to replace this with a black-box Amdahl's law demo tool is welcome.
The jargon-busting presentation was well-received.
For instructors sharing the terminal, there was a persistent problem with important commands and context scrolling off the screen when info commands were run. One learner pointed out that the Code Refinery people have a command history tool which helps with this.