Cisco XRD shell config deployment

[leec-666]

After playing with shell script config deployment for Juniper cRPD I got curious about how to do the same for some of the other container options and set my eyes on Cisco XRD.

See commit here: bfadf2c

The best approach I found is to use the ZTP command xrapply which writes and commits a file to the config:

#!/bin/bash
source /pkg/bin/ztp_helper.sh
tail -n +2 $1 > config.conf # strip first line since config is rejected when shebang is present
xrapply config.conf | tee /tmp/cli-status
if grep -i error /tmp/cli-status >/dev/null; then
  echo "Configuration load failed, aborting"
  exit 1
fi

In the limited testing I did I observed slightly shorter deployment times with netlab initial. Details in attachments:

Simple topology + Ansible: 14 seconds
Simple topology + shell script: 10 seconds
test-small.txt

Larger more complex topology + Ansible: 32 seconds
Larger more complex topology + shell script: 25 seconds
test-large.txt

I'm not sure that it saves enough time to justify its existence in the project but it was a fun exercise and I'd be interested to hear your thoughts.

[ipspace]

I dug into xrapply implementation, and it already checks for errors in output and returns the error status (which would then be propagated to bash and docker exec). Unfortunately, it looks like '-' as filename meaning stdin doesn't work :(

Looks like the minimum script that should do what we need would be

#!/bin/bash
source /pkg/bin/ztp_helper.sh
tail -n +2 $1 > /tmp/config.conf # strip first line since config is rejected when shebang is present
xrapply /tmp/config.conf

BTW, always create throwaway files in /tmp/, and if you want to make sure they're unique, use $$ in filename (not needed in this case as we're never executing multiple copies of the script in parallel)

[ipspace]

Oh, and please submit a PR -- I'm sick-and-tired of looking at Ansible noise.

[leec-666]

Thanks for the pointers. #3153.