Bug #18069
closedrados suite upgrade tests fail on xenial nodes
0%
Description
For example:
info: http://pulpito.ceph.com/teuthology-2016-11-28_11:00:15-rados-jewel-distro-basic-smithi/582105/ log: http://qa-proxy.ceph.com/teuthology/teuthology-2016-11-28_11:00:15-rados-jewel-distro-basic-smithi/582105/ sentry: http://sentry.ceph.com/sepia/teuthology/?q=eb7ca1124d2945528e3886292faff070 Failed to fetch package version from http://gitbuilder.ceph.com/ceph-deb- xenial-x86_64-basic/ref/v0.67.10/version
Presumably the Xenial gitbuilders don't have these historical tags
Updated by John Spray over 7 years ago
Similar failures when other jobs try to get infernalis branch
Updated by David Galloway over 7 years ago
Looks like somebody forced a rebuild on v0.67.10
infernalis is failing due to Error: Import No module named _markerlib
http://gitbuilder.sepia.ceph.com/gitbuilder-ceph-deb-xenial-amd64-basic/log.cgi?log=baf17c90980f3c61f75775f561ced5b2a1d2141c
Updated by Sage Weil over 7 years ago
I forced the rebuild bc (I assumed) it wasn't actually built before. I suspect the workaround here is just to not run these ancient upgrade tests on xenial, though (os_type: centos). Right now on master we only do jewel > kraken/master (hammer>jewel->x isn't merged yet). I don't think we need anything older than that?
Updated by David Galloway over 7 years ago
Sage Weil wrote:
I forced the rebuild bc (I assumed) it wasn't actually built before.
Nah, that gitbuilder never built v0.67.10. When a new gitbuilder is set up, we trick the gitbuilder into thinking all builds have passed so it doesn't go back and try to rebuild every branch. We then go through and force rebuild on branches that make sense (master, hammer, etc.)
Is it even sane to expect v0.67.10 (released in 2014) to build on Xenial (released this year)? It's failing because ceph.com doesn't listen to git requests anymore. I guess it was when v0.67.10 was released. Either way, I don't think this build failure is distro-specific. It'd fail on any gitbuilder if we forced a rebuild.
I suspect the workaround here is just to not run these ancient upgrade tests on xenial, though (os_type: centos). Right now on master we only do jewel
> kraken/master (hammer>jewel->x isn't merged yet). I don't think we need anything older than that?
Sounds right to me but not really my call.
Updated by David Galloway over 7 years ago
- Related to Bug #18089: Various official builds missing from CI/Shaman added
Updated by David Galloway over 7 years ago
distribute can be installed via easy_install but not pip.
(venv) root@gitbuilder-ceph-deb-xenial-amd64-basic:~/venv# pip install distribute Collecting distribute Downloading distribute-0.7.3.zip (145kB) 100% |████████████████████████████████| 153kB 2.8MB/s Requirement already satisfied: setuptools>=0.7 in ./lib/python2.7/site-packages (from distribute) Building wheels for collected packages: distribute Running setup.py bdist_wheel for distribute ... error Complete output from command /root/venv/bin/python2 -u -c "import setuptools, tokenize;__file__='/tmp/pip-build-XVD1VT/distribute/setup.py';f=getattr(tokenize, 'open', open)(__file__);code=f.read().replace('\r\n', '\n');f.close();exec(compile(code, __file__, 'exec'))" bdist_wheel -d /tmp/tmpMnUohqpip-wheel- --python-tag cp27: Traceback (most recent call last): File "<string>", line 1, in <module> File "/tmp/pip-build-XVD1VT/distribute/setup.py", line 58, in <module> setuptools.setup(**setup_params) File "/usr/lib/python2.7/distutils/core.py", line 137, in setup ok = dist.parse_command_line() File "setuptools/dist.py", line 276, in parse_command_line result = _Distribution.parse_command_line(self) File "/usr/lib/python2.7/distutils/dist.py", line 467, in parse_command_line args = self._parse_command_opts(parser, args) File "setuptools/dist.py", line 602, in _parse_command_opts nargs = _Distribution._parse_command_opts(self, parser, args) File "/usr/lib/python2.7/distutils/dist.py", line 523, in _parse_command_opts cmd_class = self.get_command_class(command) File "setuptools/dist.py", line 406, in get_command_class ep.require(installer=self.fetch_build_egg) File "pkg_resources.py", line 2254, in require working_set.resolve(self.dist.requires(self.extras),env,installer))) File "pkg_resources.py", line 2471, in requires dm = self._dep_map File "pkg_resources.py", line 2682, in _dep_map self.__dep_map = self._compute_dependencies() File "pkg_resources.py", line 2699, in _compute_dependencies from _markerlib import compile as compile_marker ImportError: No module named _markerlib ---------------------------------------- Failed building wheel for distribute Running setup.py clean for distribute Failed to build distribute Installing collected packages: distribute Running setup.py install for distribute ... done Successfully installed distribute-0.7.3 (venv) root@gitbuilder-ceph-deb-xenial-amd64-basic:~/venv# easy_install distribute Searching for distribute Best match: distribute 0.7.3 Adding distribute 0.7.3 to easy-install.pth file Using /root/venv/lib/python2.7/site-packages Processing dependencies for distribute Finished processing dependencies for distribute (venv) root@gitbuilder-ceph-deb-xenial-amd64-basic:~/venv# pip install distribute Requirement already satisfied: distribute in ./lib/python2.7/site-packages Requirement already satisfied: setuptools>=0.7 in ./lib/python2.7/site-packages (from distribute)
Updated by Sage Weil over 7 years ago
Yeah, I think the resolution is to just make upgardes from old versions avoid xenial. I've fixed the jewel hammer-x and clietn-upgrade tests in https://github.com/ceph/ceph-qa-suite/pull/1292.. if there are other tests just let me know, they're easy to fix!
Updated by David Galloway over 7 years ago
- Status changed from New to Resolved
- Assignee set to Sage Weil