Improve `jenkins-slave-connect.service` robustness.

Review Request #3836 — Created May 6, 2016 and submitted — Latest diff uploaded

jsirois
pants
jsirois/systemd/no_restart_rate_limits
3354
4eaab0d...
pants-reviews
patricklaw

Oft-times the service is observed failing on its 1st attempt with
/mnt/xvdb not being available. This corrects dependencies and hardens
them (both After= and Requires= are needed after reading docs more
closely) as well as un-capping retries by setting
StartLimitInterval=0.

build-support/packer/jenkins-slave.customize.sh | 14 +++++++++++---
1 file changed, 11 insertions(+), 3 deletions(-)

I have this AMI canaried on 1 instance which came up with a 1st-try
JNLP connect vs all others which were taking at least 2 tries. The CI
below went through with 2 shards running through this canary.

CI went green here:
  http://jenkins.pantsbuild.org/job/pantsbuild/job/pants/branch/PR-3354/4/
Loading...