Problem Statement
In an environment where multiple vSphere clusters are protected by VMware Site Recovery Manager (SRM) with array based replication. What is the best way to ensure the RTO/RPO is met/exceeded and to minimize the storage replication overhead?
Assumptions
1. Additional storage will not be obtained
2. Eight (8) Paths per LUN are Masked/Zoned
Motivation
1. Optimize underlying storage usage
2. Ensure transient files are not unnesasarily replicated
Architectural Decision
Configure vSphere cluster swapfile policy to Store the swapfile in the datastore specified by the host.
Create and configure a dedicated swap file datastores provided by Tier 1 storage with greater than the capacity of the total vRAM for the cluster itself, along with any/all clusters using the cluster/s as recovery sites.
Justification
1.Decreased storage replication between protected and recovery sites
2. Reduced impact to the underlying storage due to reduced replication
3. Reduces the used space at the recovery site
4. No impact to the ability to recovery to, or failback from the recovery site
5. vMotion performance will not be impacted as all hosts within a cluster share the same swap file datastore which is provided from the existing shared storage
6. There is minimal complexity in setting a dedicated swap file datastore as such, the benefits outweigh the additional complexity
7. In the event of swapping, performance will not be impacted as the swap file is on Tier 1 storage
8. There is no additional Tier 1 storage utilization as the vswap file would alternatively be set to “Store in the same director as the virtual machine”
9. Ensures memory (RAM) over commitment can still be achieved where as setting memory reservations would reduce/eliminate this benefit of vSphere
Implications
1. vMotion performance between clusters will be degraded as the swap file will be moved as part of the vMotion to the destination cluster swap file datastore
2. One (1) datastore out of a maximum of 256 per host are used for the swap file datastore
3. Eight (8) paths out of a maximum of 1024 per host are used for the swap file datastore
Alternatives
1. Store the swapfile in the same directory as the virtual machine
2. Set Virtual machine memory reservations of 100% to eliminate the vswap file
Relates Articles
1. Site Recovery Manager Deployment Location
2. VMware Site Recovery Manager, Physical or Virtual machine?
Would you eliminate implication #1 by configuring swap location on a datastore shared across all clusters?
Yes, implication #1 would be eliminated however you then need to consider if this is a good idea, and way up the pros/cons. I would suggest on the rare occasion you vMotion between clusters, the impact of moving the vSwap file can be tolerated as opposed to introducing what would be additional risk to the environment by having all clusters share a datastore for vswap.