Closed haraldschilly closed 8 years ago
Thanks for the report @haraldschilly . I thought we had this under control, but apparently not. :-(
Is there an error message above the backtrace in the log?
(pinging @mlucy )
oh sorry, I missed that there is some part above! That looks indeed helpful for you:
2016-03-30T08:42:13.182277319 2773091.448582s info: Table c99153c0-432d-40f0-8626-f0a24520dc72: Configuration is changing.
2016-03-30T08:45:57.948470411 2773316.214775s info: Table 05c83381-98cf-4c3d-af1b-1cbe1a3e4b69: Configuration is changing.
2016-03-30T08:46:00.494933067 2773318.761238s info: Table 4d50504e-a067-4979-ae80-3ca335f489b0: Configuration is changing.
2016-03-30T08:46:14.222710664 2773332.489015s info: Table 05c83381-98cf-4c3d-af1b-1cbe1a3e4b69: Configuration is changing.
2016-03-30T08:47:29.684079835 2773407.950385s error: Error in src/rdb_protocol/store.cc at line 960:
2016-03-30T08:47:29.685241407 2773407.951546s error: Guarantee failed: [!pair.first.inner.overlaps(region.inner)]
2016-03-30T08:47:29.685273472 2773407.951578s error: Backtrace:
2016-03-30T08:47:29.686709725 2773407.953014s error: Error in src/rdb_protocol/store.cc at line 960:
2016-03-30T08:47:29.686730414 2773407.953035s error: Guarantee failed: [!pair.first.inner.overlaps(region.inner)]
2016-03-30T08:47:29.686746154 2773407.953050s error: Backtrace:
Thanks, that error should indeed be helpful.
A potential fix for this is in next
(OTS by @danielmewes). We think the bug might be caused by calling note_reshard
in the destructor of local_replicator_t
, which previously happened before that object's auto-drainer was destroyed, meaning that changefeed_subscribe_t
s on the old region might still be queued up and wind up happening after note_reshard
.
I'm going to close this for book-keeping reasons. If we see the issue again after RethinkDB 2.3.0, we will re-open it.
2 out of 6 nodes crashed when rebalancing. minutes later everyone is happy again, but that's still bad. I found one stacktrace, that might be helpful...
version is 2.5.5