Description
Scenario:
1. Sync Gateway creates gocb client connected to a 2 node Server cluster (node A and B) running 4.62
2. Node A is rebalanced out
3. Node A is upgraded to 5.0, restarted
4. Node A is rebalanced in
5. Node B is rebalanced out
6. Node B is upgraded to 5.0, restarted
7. Node B is rebalanced in
We're seeing the following errors in Sync Gateway when trying to perform kv operations via this process:
a. After step 4, all kv operations return a 'not connected to a bucket' error.
b. After step 5, all kv operations return a 'no access' error.
The 'no access' errors continue until the gocb instance is dropped and recreated (in our case, Sync Gateway is restarted).
Should gocb be handling this upgrade scenario internally, or are consumers expected to detect these errors and drop/re-establish connections manually?
Attachments
Issue Links
For Gerrit Dashboard: GOCBC-235 | ||||||
---|---|---|---|---|---|---|
# | Subject | Branch | Project | Status | CR | V |
83868,3 | GOCBC-235: Always SELECT_BUCKET on Spock clusters. | v7 | gocbcore | Status: MERGED | +2 | +1 |