Fail Over an EMC ViPR Site
Table of Contents
It includes a sample geo configuration, the behavior of ViPR Controller and ViPR Data Services during the failure, and what steps are needed for recovery.Back to Top
- Site 1 running ViPR Controller and ViPR Data Services
- Site 2 running ViPR Controller and ViPR Data Services
- Site 3 running ViPR Controller and ViPR Data Services
- Application 1 in New York, has local affinity to Site 1 New York
- Application 2 in Los Angeles, has local affinity to Site 2 Los Angeles
Before any corrective step is taken by the administrator, the behavior at this point is as follows:
- ViPR Data Services:
- For buckets created in Sites 2 and 3: all operations succeed.
- For buckets created in Site 1 (the failed site)
- Reads and writes fail at any site.
- Reads fail for data created in Site 1.
- Reads succeed for data created in Site 2 or 3.
- New writes succeed if name check is not selected, but byte-range updates fail.
- ViPR Controller:
- Global resources are in read-only mode across the geo configuration, but provisioning is possible at Site 2 and Site 3. New writes to sites 2 and 3 succeed.
Before you begin
Ensure that Site 1 is not processing any requests. This can be done by disabling the network.
- Log in to the UI with an account that has the System Administrator role (but not root user).
- In Admin mode, select Virtual Assets > Object Virtual Pools.
- Select an object virtual pool that is used by the VDC at the failed site.
- Click Remove next to the failed VDC (see screenshot below). This operation tells ViPR that the failed VDC is no longer part of the multisite configuration.
- Confirm that you want to remove the VDC.
- Repeat steps 3 and 4 for each object virtual pool used by the VDC at the failed site
After failover, data services are operational on Site 2 and Site 3.
During the rebalancing process ViPR cannot tolerate another site failure, and you need to closely watch the rebalancing process and ensure it completes. The rebalancing process depends on the amount of data being replicated and the network performance.Back to Top
- Restore the ViPR Controller from backups. You can find details in EMC ViPR native backup and restore service.
- Run the ViPR Controller on the restored site.
- On the restored site, delete the virtual data center and all resources. (ViPR Data Services cannot be deployed to a site with the same ID.)
- Redeploy the ViPR Controller on the recovered site.
- Redeploy ViPR Data Services on the recovered site.