@procogent Those messages make sense. `rexi_DOWN` and `mem3_rpc` are expected because one of the nodes was down during the upgrade for a short period of time and internal replicator wasn't able to talk to that node. That's ok because it will keep retrying unit it finally is able to connect.
`[<<"couch_stream:foldl/6 L91">>,<<"couch_att:to_binary/2 L647">>,..` error is from attachment streaming. It turns out that even even without this commit https://github.com/apache/couchdb/commit/56782453f342fb5e4137e8c9afc79b1992a8b21a attachment upload would not be working when upgrading from 2.1.1 to 2.2. Notice that since 2.1.1 there was already a change in https://github.com/apache/couchdb/commit/0e926881c2c41a557cac035a779e2a95d13168ca#diff-b23fc0c1115730cb26590a6db318107e which modified the `fabric_doc_attachments.erl` module. That's exactly the problem with sending function closure between nodes - that any change in the module where the code lives in a cluster will break and start throwing errors. [ Full content available at: https://github.com/apache/couchdb/issues/1578 ] This message was relayed via gitbox.apache.org for [email protected]
