Scap deploying arbitrary revision, incorrectly reporting correct revision
Closed, ResolvedPublic
Actions

Assigned To

Authored By

	awight
	Nov 7 2017, 6:24 PM

Description

Steps to reproduce

I'm on beta tin, and have checked out revision rORESDEPLOY29905e509984 into /srv/deployment/ores/deploy.
scap deploy -l "ores1*" "test deployment to ores* (non-production)"

P6281 (An Untitled Masterwork)

1	18:15:04 [tin] Started deploy [ores/deploy@29905e5]
2	18:15:04 [tin] Started deploy [ores/deploy@29905e5]: test deployment to ores* (non-production)
3	18:15:04 [tin]
4	== CLUSTER ==
5	:* ores1001.eqiad.wmnet
6	:* ores1003.eqiad.wmnet
7	:* ores1005.eqiad.wmnet
8	:* ores1007.eqiad.wmnet
9	:* ores1002.eqiad.wmnet
10	:* ores1008.eqiad.wmnet
11	:* ores1009.eqiad.wmnet
12	:* ores1004.eqiad.wmnet
13	:* ores1006.eqiad.wmnet
14	18:15:05 [ores1001.eqiad.wmnet] Fetch from: http://tin.eqiad.wmnet/ores/deploy/.git
15	18:15:07 [ores1001.eqiad.wmnet] Checkout rev: 4f2d1d0389449cedb26cc7f5f31ffab2098cb203
16	18:15:07 [ores1001.eqiad.wmnet] Unhandled error:
17	Traceback (most recent call last):
18	File "/usr/lib/python2.7/dist-packages/scap/cli.py", line 308, in run
19	exit_status = app.main(app.extra_arguments)
20	File "/usr/lib/python2.7/dist-packages/scap/deploy.py", line 139, in main
21	getattr(self, stage)()
22	File "/usr/lib/python2.7/dist-packages/scap/deploy.py", line 299, in fetch
23	git.checkout(rev_dir, self.rev)
24	File "/usr/lib/python2.7/dist-packages/scap/git.py", line 350, in checkout
25	subprocess.check_call(cmd, shell=True)
26	File "/usr/lib/python2.7/subprocess.py", line 540, in check_call
27	raise CalledProcessError(retcode, cmd)
28	CalledProcessError: Command '/usr/bin/git checkout --force --quiet 4f2d1d0389449cedb26cc7f5f31ffab2098cb203' returned non-zero exit status 128
29	18:15:07 [ores1001.eqiad.wmnet] deploy-local failed: <CalledProcessError> {u'cmd': u'/usr/bin/git checkout --force --quiet 4f2d1d0389449cedb26cc7f5f31ffab2098cb203', u'output': None, u'returncode': 128}
30	18:15:07 [tin] [u'/usr/bin/scap', u'deploy-local', u'-v', u'--repo', u'ores/deploy', u'-g', u'cluster', u'fetch', u'--refresh-config'] on ores1001.eqiad.wmnet returned [70]: http://tin.eqiad.wmnet/ores/deploy/.git
31	From http://tin.eqiad.wmnet/ores/deploy/
32	* [new branch] CELERY_4 -> origin/CELERY_4
33	* [new branch] STABLE_REVSCORING_1 -> origin/STABLE_REVSCORING_1
34	* [new branch] master -> origin/master
35	* [new tag] scap/sync/2017-11-06/0009 -> scap/sync/2017-11-06/0009
36	Cloning into '/srv/deployment/ores/deploy-cache/revs/4f2d1d0389449cedb26cc7f5f31ffab2098cb203'...
37	done.
38	fatal: reference is not a tree: 4f2d1d0389449cedb26cc7f5f31ffab2098cb203
39
40	18:15:07 [tin] 1 targets had deploy errors

It actually tries to checkout an ancient revision. In this case it fails (inexplicably), but I've also seen it succeed in deploying the wrong revision, all the while reporting the correct one in the logs.

As a workaround, I've been including "-r HEAD" in my commandline arguments.