PowerScale: Create a cascaded replication properly on OneFS
Summary: Below are steps to properly configure cascaded replication between Cluster_A, Cluster_B, and Cluster_C.
Instructions
The main idea to keep in mind while configuring a cascaded replication. Ensure that replication job A_B should finish with "success/needs_attention" before starting replication job B_C.
Always, we should avoid that replication job A_B status is "failed OR canceled OR running OR paused" while replication job B_C starts.
In order to reach the above conditions:
1. We can make the replication policy from A_B to create a target snapshot with the name pattern "Archive-%{SrcCluster}-%{PolicyName}-%Y-%m-%d_%H-%M-%S" when the job is finished with success.Cluster_A# isi sync policies create --name=A_B --source-root-path=/ifs/folder_A --target-host=xxx.xxx.xxx.xxx --target-path=/ifs/folder_B --action=sync --target-snapshot-archive=yes --target-snapshot-pattern=Archive-%{SrcCluster}-%{PolicyName}-%Y-%m-%d_%H-%M-%S --target-snapshot-expiration=5D
--target-snapshot-pattern=Archive-%{SrcCluster}-%
- Then we make the replication policy from B_C to start only when a source snapshot pattern "Archive-*" is created.
Cluster_B# isi sync policies create --name=B_C --source-root-path=/ifs/folder_B --target-host=xxx.xxx.xxx.xxx --target-path=/ifs/folder_C --action=sync --schedule=when-snapshot-taken --snapshot-sync-pattern="Archive-*" --snapshot-sync-existing=yes
This will guarantee that the B_C replication job starts only after A_B replication job completes with success + the B_C schedule is tied to the A_B schedule.
Check below from LAB:Cluster_A# cd /ifs/folder_ACluster_A# ls -lhtotal 10249-rw-r--r-- 1 root wheel 10M Jul 14 20:51 fileCluster_A# isi sync jobs start A_BCluster_A# isi sync reports listPolicy Name Job ID Start Time End Time Action State-----------------------------------------------------------------------------A_B 1 2021-07-14T20:51:22 2021-07-14T20:51:31 run finished-----------------------------------------------------------------------------Total: 1Cluster_B# cd /ifs/folder_BCluster_B# ls -lhtotal 10249-rw-r--r-- 1 root wheel 10M Jul 14 20:51 fileCluster_B# isi sync reports listPolicy Name Job ID Start Time End Time Action State-----------------------------------------------------------------------------B_C 2 2021-07-14T20:51:51 2021-07-14T20:52:10 run finishedB_C 1 2021-07-14T20:51:32 2021-07-14T20:51:38 run finished-----------------------------------------------------------------------------Total: 2Cluster_C# cd /ifs/folder_CCluster_C# ls -lhtotal 10249-rw-r--r-- 1 root wheel 10M Jul 14 20:51 file
- The cascaded replication is one-way replication and right in specific workflows and in migrations.
- Failover and Failback is ONLY SUPPORTED between Cluster A and Cluster B. Failover between B and C is NOT SUPPORTED due to target overlap on cluster B.
- On Failover to cluster B, the schedule must be modified to support replication from B back to A.