[ODLPARENT-135] Karaf feature:installation "deadlock" Created: 13/Dec/17 Updated: 15/Dec/17 Resolved: 15/Dec/17 |
|
| Status: | Resolved |
| Project: | odlparent |
| Component/s: | Karaf |
| Affects Version/s: | 3.0.2 |
| Fix Version/s: | 3.0.2 |
| Type: | Bug | Priority: | Medium |
| Reporter: | Michael Vorburger | Assignee: | Stephen Kitt |
| Resolution: | Cannot Reproduce | Votes: | 0 |
| Labels: | None | ||
| Remaining Estimate: | Not Specified | ||
| Time Spent: | Not Specified | ||
| Original Estimate: | Not Specified | ||
| Attachments: |
karaf-feature-install-deadlock-jstack.txt
karaf-feature-install-deadlock-jstack_20-minutes-minutes-later.txt
karaf-feature-install-deadlock-jstack_5-minutes-minutes-later.txt
|
| Description |
|
When attempting to bump the DAEXIM project to the new odlparent (3.0.x) and yangtools (2.0.x) following https://wiki.opendaylight.org/view/ODL_Root_Parent:BigBumpOfJan2018, as in https://git.opendaylight.org/gerrit/#/c/66431/ (which right now is JUST some POM version bumps, no other changes), then SFT fails with java.lang.ClassNotFoundException: org.junit.runners.model.TestTimedOutException in features/odl-daexim. That exception just means Pax Exam timed out waiting for Karaf (although ideally there should not be a ClassNotFoundException; but that's a minor problem). If we build the features etc. with -Pq and try it out manually via feature:install odl-daexim then we see that never returns and Karaf spin looping around something (not, technically, dead locked); attached x3 jstack illustrating where it seems to have gotten stuck. There is nothing at all in data/log/karaf.log. This doesn't provide anything more of interest either: karaf@root()> feature:install --verbose --no-auto-refresh --all-wiring --store resolution.txt --simulate odl-daexim Adding features: odl-daexim/[1.3.0.SNAPSHOT,1.3.0.SNAPSHOT] FYI: That features/odl-daexim is the one without even restconf, just real basics. (DAEXIM has a restconf-all feature which also includes RESTCONF.) |
| Comments |
| Comment by Stephen Kitt [ 14/Dec/17 ] |
|
Just a quick comment re the last test mentioned: --simulate is rather useless in Karaf, you really need to perform a real feature:install -v to get anything useful. |
| Comment by Stephen Kitt [ 15/Dec/17 ] |
|
Looking into this further, how do you even get the features built? The build fails for me in daexim-test-model with a YANG error. |
| Comment by Stephen Kitt [ 15/Dec/17 ] |
|
(The error is easy to fix, just rename the files to use correct revisions.) |
| Comment by Stephen Kitt [ 15/Dec/17 ] |
|
Once the YANG and Checkstyle issues are fixed, daexim builds and tests fine. The original bug report seems indicative of a build with mixed yangtools, e.g. one with an unbumped mdsal. |