Cross-Region Replication FAQ
How can I verify if data is replicated between primary and Disaster Recovery (DR) clusters?
You can run
gstatusgraph from the Linux terminal on both the primary and DR cluster.
The count for vertices and edges should match if data is in sync.
Note that if there are running loading jobs, the DR cluster might show a lower count.
In this case, check again when the loading job is done.
Why am I not seeing any loading job declared in the DR cluster?
Loading jobs themselves are not replicated to the DR cluster. However, the data loaded by these loading jobs is replicated.
What happens after I run
DROP ALL on primary?.
DROP ALL command will stop cross-region replication (CRR).
You will need to restore to re-establish the feature again.
Here is a list of all commands and operations that will stop CRR:
gsql drop allwhich clears all data and schema
gsql clear graph storewhich clears only data
gsql --resetwhich clears all data, schemas and users, even resetting the password of the default tigergraph.
gsql import graph
gsql export graph
Why is GSQL failing to replay a replica with an "UNAUTHORIZED" error?
It’s most likely that primary and DR have different passwords for the same TigerGraph user.
This can happen when you enable CRR without restoring the GBAR backup in the DR cluster (since you did not have any data), but the DR was installed with a different password than the primary. Make sure the DR and primary clusters have the same TigerGraph password before enabling CRR.
What happens if DR is down, unavailable or under scheduled maintenance (e.g. VM Motion)?
Nothing will happen.
As soon as DR is back online, Kafka MirrorMaker will replicate the Kafka Topic and GSQL will start replaying the replicas from where it left off.
in order to DR automatically recover, it has to come back up within the Kafka Topic retention time limit.
By default, this is set to 168 hours (7 days).
You can tune this parameter based on your need by running
gadmin config set Kafka.RetentionHours <value_in_hours>
Can I have multiple DRs?
Yes. You can set up multiple DR clusters to sync with one primary cluster. There is no hard limit to the number of DR clusters.
How will my application write to the new primary after DR failover?
We suggest that you handle this with an application load balancer where you can configure the DR IP hosts list (e.g. if you are using NGINX you can add the DR hosts list in the upstream section). When the Load Balancer fails the health check on the current primary it will re-route the traffic to the DR host list. You should then manually fail over to the DR cluster.