Now that your cloud is up and running, your team receives an emailed weekly capacity report from email@example.com. Here is how you read it and plan for adding capacity.
This section contains the information for your installation, the size of it, as well as how many VCPUs are reserved. Typically your reservations will not exceed 70-75%, so watch for large percentage increases in instance launches month-over-month.
If you need to add capacity, you may want to add an availability zone (AZ). Each AZ is independent from other zones, with new API endpoints and a different dashboard URL. Availability zones give you physical isolation and redundancy of resources, with separate power supplies and networking equipment, so that users can achieve high availability if one zone becomes unavailable.
Storage Usage (only with external storage)
When you have an external storage solution, you have a storage usage section. It indicates the size, how much is used and how much is available. These are labeled by file system. When you run out of physical storage (over 70% raw capacity). Metapod support opens a ticket when levels hit pre-defined thresholds.
Volume Allocation (only with external storage)
The Volume Allocation section indicates when your storage is over subscribed, by how much, and indicates available volumes. So you may see a larger value in the raw storage column than in the available column. For example, with 2.0 TB of raw storage, you can have 3.0 TB available when oversubscribed by 1.5 times.
Flavor Types In Use
Even with external storage, your flavors can have ephemeral storage or external storage. This section indicates which flavors are in use, along with name, memory amount, VCPUs, how much storage the root takes up, and how much ephemeral storage is available on that flavor.
Instance Counts By Flavor
By tracking which instance flavors are most launched, you can trim down the number of flavors you offer as needed. This section of the report gives percentage increases or decreases as well as showing when there’s no change. Watch for variances depending on what’s needed by your users, seasonal needs, and which guest operating systems are required.
Look for Warnings here for particular types of instances. Those instances cannot be launched until the RAM or disk available is available to those instances.
Detailed node breakdown
At the bottom of this table there’s a legend indicating what each column means. Here’s a copy of the legend for interpretation.
Some of the reported values are calculated through settings in
ram_allocation_ratio setting configures a virtual RAM to physical RAM allocation ratio which affects all RAM filters upon scheduling where new instances are launched. Same for
disk_allocation_ratio, these are also accounted for when scheduling instance launches. The
reserved_host_memory_mb value is 512 by default, and affects the effective memory available to the virtual machines.
Contact support if you want to further tune those values for your environment.
- TCPU - Total VCPUs on the host accounts for
- ACPU - Allocated VCPUs on the host
- FCPU - Free VCPUs on the host
- TRAM - Total RAM on the host
- URAM - Total usable RAM on the host accounts for
- ARAM - Total allocated RAM on the host
- FRAM - Total free RAM for scheduling on the host
- TDSK - Total disk on the host
- UDSK - Total usable Disk on the host accounts for
- ADSK - Allocated disk on the host
- FDSK - Free disk on the host
You may also look at the bottom of the report email to find where there are instances with older container versions on a particular node, those nodes/hypervisors need to be rebooted.