Balajee Venkatesh
Balajee Venkatesh

Reputation: 1099

Bigtable replication guidance

As per the Bigtable documentation:

Bigtable allows read/write operations to be performed on the basis of shortest distance.

Hence, it should always write to or read from nearest available cluster, in order to validate this, I had set up a GCE instance in "us-east1-d" zone and created a Bigtable instance with two clusters where one was in "us-east1-d" zone and other was in "us-west1-a"zone.

I expected my writes from GCE instance to follow "us-east1-d" only but I noticed a few writes going into "us-west1-a" and then replicating into "us-east1-d". I tracked this behaviour from Bigtable replication monitoring console only. Keeping in mind that zonal unavailability/disruption might be the reason of this behaviour, I deleted and recreated the Bigtable instance with clusters in different regions and zones. I still see some varience from "shortest distance" read/write theory of Bigtable.

So, my question, is there any way which can help me verify that read/writes from east region will always go to east region provided there is no zonal failure? Reliability of this feature is extremely critical to my applications in order to facilitate lowest possible latency.

Upvotes: 0

Views: 642

Answers (1)

Rafael Lemos
Rafael Lemos

Reputation: 5829

You could use Cloud Monitoring for BigTable to check each instance specifically, is not exactly what you want, since it's not per region and/or zone, but could be helpful to analize incoming requests to figure out what is happening with your particular instance.

One metric that is available and could really be helpful in your case is the Automatic failovers metric which shows the number of requests that were automatically rerouted from one cluster to another due to a failover scenario, such as a brief outage or delay.

You can see more details on how to set it up in this documentation.

Hope this helps.

Upvotes: 2

Related Questions