Claiming too quickly after an unclaim causes 'error: lock instance is no longer acquired'

[topherbullock]

We have two jobs which try to acquire a lock on the same pool ( with only one lock to acquire ), the first one to acquire the lock failed to release the lock because the second claimed the env immediately after the unclaim commit. Probably related to this check

Job 1

:arrow_up: Env

  acquiring lock on: aws-bosh-release

:arrow_down: Env

  Cloning into '/tmp/build/get'...
  9f3afaf claiming: mars

bunch of jobs and stuff

:arrow_up: Env

releasing lock: mars on pool: aws-bosh-release 

:arrow_down: Env

  Cloning into '/tmp/build/get'...
  f1169b2 unclaiming: mars
  error: lock instance is no longer acquired

/

Job 2 :arrow_up: Env

  acquiring lock on: aws-bosh-release
  .........................................................

:arrow_down: Env

  Cloning into '/tmp/build/get'...
  70c75d1 claiming: mars

Commits in Git logs show that Job #2's claim happened immediately after Job #1 's unclaim.

[concourse-bot]

Hi there!

We use Pivotal Tracker to provide visibility into what our team is working on. A story for this issue has been automatically created.

The current status is as follows:

• [x] #100743632 Claiming too quickly after an unclaim causes 'error: lock instance is no longer acquired'

This comment, as well as the labels on the issue, will be automatically updated as the status in Tracker changes.

[vito]

sweet emoji art

[jtarchie]

Reproducible steps:

Setup

mkdir -p repo/pool/unclaimed
mkdir -p repo/pool/claimed
cd repo
git init
git add -A
git ci -m 'init pool of resources'

Job A performing the put on a concourse pipeline

git mv pool/unclaimed/RESOURCE pool/claimed/RESOURCE
git ci -m 'claiming resource from Job A'

git mv pool/claimed/RESOURCE pool/unclaimed/RESOURCE
git ci -m 'unclaiming resource from Job A'
ref=$(git log HEAD --format="%H" | head -1)

Job B claiming the resource

git mv pool/unclaimed/RESOURCE pool/claimed/RESOURCE
git ci -m 'claiming resource from Job B'

Job A then running the /opt/resource/in script on it's ref

$ git log --oneline $ref..HEAD -- pool/
869df45 claiming resource from Job B

The following is showing the error message from here.

It looks like a possible solution may be just filtering out subsequent claimed messages.

For example.

$ git log --grep '^claiming' --invert-grep --oneline $ref..HEAD -- pool/

There is no output from this command.

[jtarchie]

We've tested the above fix on our concourse deployment and we no longer experience the error.

[concourse-bot]

Hello again!

All stories related to this issue have been accepted, so I'm going to automatically close this issue.

At the time of writing, the following stories have been accepted:

• #100743632 Claiming too quickly after an unclaim causes 'error: lock instance is no longer acquired'

If you feel there is still more to be done or have any questions, feel free to reopen!