vSAN Best Practices for Clusters Using Large Capacity Tier Disks
Article ID: 326699
Updated On:
Products
VMware vSAN
Issue/Introduction
As customers start to consider using larger capacity drives greater than 8TB per host on vSAN, they need to take into consideration the following caveats, trade-offs, and best practices.
Environment
VMware vSAN 7.0.x
Resolution
Caveats and Trade-Offs
One of the biggest caveats of using larger capacity tier disks per host is the increase in resync duration upon drive or host failure. The increase in resync duration is typically proportional to the capacity being rebuilt. Therefore, in the event of a failure on large capacity disks, the system will take proportionately longer to get back into compliance due to the longer resync duration.
To mitigate resync duration and its impact, we recommend the following best practices for clusters using large capacity tier disks i.e., deployments with ~ 100+ TB per host.
Best Practices
a. Recommend using FTT=2 over FTT=1 for vSAN clusters. (please note FTT=2 requires a minimum of 5 hosts in a cluster)
Given that resync durations are longer during a failure scenario in a large capacity configuration, we recommend using FTT=2 over FTT=1 so that the cluster has higher availability (and durability) during the longer resync period.
b. Recommend turning off deduplication
Given that the fault domain in a system with Deduplication enabled is larger (disk group), the entire capacity of the disk group needs to be resynced in the event of a single drive failure. Therefore, this increases the amount of data and consequently the amount of time it takes to resync. Furthermore, commonly deployed large capacity OLTP databases benefit more from compression-only for space savings since databases typically compress well but do not deduplicate very well. Therefore, Space efficiency can be achieved more efficiently by using compression only for large capacity OLTP (Online Transaction Processing) databases.
c. Recommend using >= 25Gbps networking
To ensure that the network bandwidth is not saturated during long resync operations, we recommend using > 25Gbps networking with larger-capacity nodes.
One of the biggest caveats of using larger capacity tier disks per host is the increase in resync duration upon drive or host failure. The increase in resync duration is typically proportional to the capacity being rebuilt. Therefore, in the event of a failure on large capacity disks, the system will take proportionately longer to get back into compliance due to the longer resync duration.
To mitigate resync duration and its impact, we recommend the following best practices for clusters using large capacity tier disks i.e., deployments with ~ 100+ TB per host.
Best Practices
a. Recommend using FTT=2 over FTT=1 for vSAN clusters. (please note FTT=2 requires a minimum of 5 hosts in a cluster)
Given that resync durations are longer during a failure scenario in a large capacity configuration, we recommend using FTT=2 over FTT=1 so that the cluster has higher availability (and durability) during the longer resync period.
b. Recommend turning off deduplication
Given that the fault domain in a system with Deduplication enabled is larger (disk group), the entire capacity of the disk group needs to be resynced in the event of a single drive failure. Therefore, this increases the amount of data and consequently the amount of time it takes to resync. Furthermore, commonly deployed large capacity OLTP databases benefit more from compression-only for space savings since databases typically compress well but do not deduplicate very well. Therefore, Space efficiency can be achieved more efficiently by using compression only for large capacity OLTP (Online Transaction Processing) databases.
c. Recommend using >= 25Gbps networking
To ensure that the network bandwidth is not saturated during long resync operations, we recommend using > 25Gbps networking with larger-capacity nodes.
Additional Information
Feedback
Yes
No
Powered by
