Details
-
Bug
-
Resolution: Fixed
-
Critical
-
Mad Hatter
-
1
Description
Need user recommendations for the following:
1. Timeout value during sync_writes for different durability_levels since they are going to take different time to serve the request
Scenario: (Consider 4 node cluster with bucket replicas=2)
|
1. Perform sync_write with timeout=3 and durability=MAJORITY
|
2. Perform sync_write with timeout=3 and durability=MAJORITY_AND_PERSIST_ON_MASTER
|
3. Perform sync_write with timeout=3 and durability=PERSIST_TO_MAJORITY
|
|
In the above three cases, step-3 can fail with minimum supported timeout value because the 'PERSIST_TO_MAJORITY' needs more time than the other 2 to satisfy the durability even in best-case (no network latency / no server load) scenario.
|
|
We need to inform the user about this and should avoid giving same values for all three Durability levels we support currently.
|
2. Batch size during bulk doc inserts / upserts with Sync_Writes
Scenario:
|
1. Perform bulk set / insert for 100 docs with durability=MAJORITY, timeout=5 seconds (Succeeds)
|
2. Now user performs same operation with say 10000 docs, last few requests will timeout in the sdk_client side even before the server actually process the requests. (Similar to MB-34576)
|