Details
-
Bug
-
Resolution: Fixed
-
Blocker
-
.master
-
Security Level: Public
-
Untriaged
-
2
-
June 30 - July 18
Description
The UPR Stream Request doesn't return the correct result, unless it's restarted. Here's how to reproduce it.
Apply the testrunner patch. That is a testcase that does the rebalance (I got it as minimal as possible).
Put the attached dev-2-nodes.ini into the testrunner/b/resources/ dirctory.
Pyt the attached rebalance_bug.py into the pyupr directory.
Now start a 2 node cluster with ./cluster_run and 8 partitions/vBuckets:
COUCHBASE_NUM_VBUCKETS=8 ./cluster_run -n 2
Run the new testrunner test (from the testrunner directory) via:
./testrunner -i b/resources/dev-2-nodes.ini -t view.createdeleteview.CreateDeleteViewTests.rebalance_breaks_upr_stream,ddoc_ops=delete,test_with_view=True,nodes_in=3,num_ddocs=2,num_views_per_ddoc=3,items=2000
Kill the test when it shows prints out:
Now run the rebalance_bug.py from your pyupr directory. The output should be something like:
the number of total changes I got from partition 4 is 250
the number of total changes I got from partition 5 is 409
the number of total changes I got from partition 6 is 296
the number of total changes I got from partition 7 is 438
Now kill the cluster and restart it.
When you now run the rebalance_bug.py script it returns the expected result:
the number of total changes I got from partition 4 is 250
the number of total changes I got from partition 5 is 250
the number of total changes I got from partition 6 is 250
the number of total changes I got from partition 7 is 250
Attachments
Issue Links
- is triggering
-
MB-10472 Handle multiple snapshots properly in view engine
- Resolved