Introduce a new transaction type for reading from secondary replicas #110
Conversation
protobuf/transaction.proto
Outdated
| @@ -67,6 +67,7 @@ message Transaction { | |||
| enum Type { | |||
| READ = 0; | |||
| WRITE = 1; | |||
| READ_SECONDARY = 2; | |||
There was a problem hiding this comment.
Defining cluster specific type here is not ideal, and I've created an issue for it: #113
There was a problem hiding this comment.
Ah I've answered it here on the other PR: https://github.com/graknlabs/client-java/pull/243/files#r563711191
But let's continue the discussion about the name here so that it's not scattered.
The reasoning to rename it to READ_SECONDARY is because it allows you to read from secondary replicas, whereas READ and WRITE reads to the primary replica (ie., the term replica does not only refer only to the secondary one).
The alternative name would be READ_SECONDARY_REPLICA?
There was a problem hiding this comment.
Interesting. I can see the logic.
I'm not familiar enough with this domain to know what would/wouldn't be an appropriate name. READ_SECONDARY_REPLICA sounds so damn long. I therefore propose that we go with READ_SECONDARY as you suggest and maybe revisit this after the first release.
There was a problem hiding this comment.
After chatting to Haikal we agreed to change READ_SECONDARY from being a Transaction type to being an Option named SECONDARY_REPLICA or similar
lolski
force-pushed
the
read-secondary
branch
from
964ef52 to
68cb20d
Compare
protobuf/options.proto
Outdated
| @@ -44,4 +44,7 @@ message Options { | |||
| oneof schema_lock_acquire_timeout_opt { | |||
| int32 schema_lock_acquire_timeout_millis = 6; | |||
| } | |||
| oneof primary_replica_opt { | |||
| bool primary_replica = 7; | |||
There was a problem hiding this comment.
I wonder if primary_replica is misleading because it could be switched off and the primary replica might still be used. What do you think of the following alternatives:
require_primary_replicaallow_secondary_replica
Or do you think we should stick with primary_replica for conciseness?
What is the goal of this PR?
We've introduced a new Grakn-Cluster-specific transaction type, "read secondary". It is a read-only transaction that may read not only from primary but also from secondary replicas.
What are the changes implemented in this PR?
allow_secondary_replica