Details
-
Bug
-
Resolution: Fixed
-
Critical
-
5.0.0
Description
Our longevity test has been going for 7 days. The test runs our integration suite, except with ephemeral buckets:
https://github.com/couchbaselabs/sequoia/blob/master/tests/integration/README.md
At the beginning of today's cycle the test is hanging on an index build on the CUSTOMER bucket. I saw these errors refers to response from node .83:
2017-08-22T11:45:49.705-07:00 [Error] Indexer::handleInitialBuildDone Stream INIT_STREAM Bucket CUSTOMER Error from Projector proto: can't skip unknown wire type 6 for protobuf.TimestampResponse. Retrying. |
2017-08-22T11:47:50.014-07:00 [Error] KVSender::addIndexForExistingBucket MAINT_STREAM CUSTOMER Error Received proto: can't skip unknown wire type 6 for protobuf.TimestampResponse from 172.23.105.83:9999 |
|
|
I checked the projector logs on .83 and saw there is an error about the bucket being invalid
2017-08-22T11:48:51.000-07:00 [Error] FEED[<=>MAINT_STREAM_TOPIC_d533f95b36e4ab615baf41b640b25f75(127.0.0.1:8091)] ##94d addInstances() invalid-bucket "CUSTOMER" |
There are other logs on this node (.83) relating to the default bucket having "Metadata overhead warning. Over 291696004715%" (which is a separate bug) but could indicate some instability on the index node.