Resources and Sizing on vanilla Openstack
Resources from heat template
TODO: update for Nov 2017 R1
We have setup an instance of ONAP based on "vanilla" Openstack (Ocata). Similarly to Resources and Sizing on Rackspace I would like to have a space for vanilla Openstack installs. The table below is generated by a little script (see attachment) based on the flavors used in our setup and the onap_openstack.yaml file.
Nr. | VM | FLAVOR | CPU | RAM | DISK |
1 | dns | SMALL | 1 | 2 | 20 |
2 | aai | XLARGE | 8 | 16 | 160 |
3 | mso | LARGE | 4 | 8 | 80 |
4 | mrouter | LARGE | 4 | 8 | 80 |
5 | robot | SMALL | 1 | 2 | 20 |
6 | vid | MEDIUM | 2 | 4 | 40 |
7 | sdnc | LARGE | 4 | 8 | 80 |
8 | sdc | XLARGE | 8 | 16 | 160 |
9 | portal | LARGE | 4 | 8 | 80 |
10 | dcae_c | MEDIUM | 2 | 4 | 40 |
11 | policy | XLARGE | 8 | 16 | 160 |
12 | appc | LARGE | 4 | 8 | 80 |
total | ..... | 50 | 100 | 1000 |
(n.b.: Need to update #vCPU for dcae still)
Openstack Defaults.
To get the HEAT template to load, we needed to change some defaults in the Openstack install.
I think we increased RAM, Instances, vPCUs to enable the resources needed by the HEAT template
Limit | |
---|---|
Injected File Content Bytes | 10240 |
Metadata Items | 128 |
Server Group Members | 10 |
Server Groups | 10 |
RAM (MB) | 128000 |
Key Pairs | 100 |
Length of Injected File Path | 255 |
Instances | 20 |
Injected Files | 5 |
VCPUs | 64 |
Per Volume Size (GiB) | -1 |
Total Size of Volumes and Snapshots (GiB) | 1000 |
Backup Size (GiB) | 1000 |
Volume Snapshots | 10 |
Volumes | 15 |
Backups | 10 |
Physical infrastructure
After we discovered a performance problem on our first lab infrastructure, we tried to setup Openstack & ONAP on a bare metal server in Softlayer. This was the break through which let a lot of issues we had before simply disappear. Nevertheless, some of the experience made it into a JIRA entry to improve the way the VMs and/or containers are started.
Role | CPU Type | cores | Disk (Gb) | RAM (Gb) | Remarks |
Compute | Dual processor E5 2690 v4 | 56 | 600GB SAS (RAID) | 128 | |
Compute | Dual processor E5 2690 v4 | 56 | 600GB SAS (RAID) | 128 | |
Controller | 4 CPU (VM) | 100 | 12 | Virtual instance on a host distinct from compute nodes |
The disk size is a bit small, we decreased the disk sizes in the big flavors (a see could see that from 160GB requested disk size only ~6GB were used after install). We will change compute nodes shortly to ones with more disk space.
Other observations
A few observations we made during the setup of on vanilla Openstack.
Robot_vm
The demo.sh script fails all tests, with the first being "access". Further investigation unveils that a file contains a parameters pointing to Rackspace.
/opt/eteshare/config/integration_robot_properties.py , at or around line 30 "GLOBAL_OPENSTACK_KEYSTONE_SERVER" to be changed to your keystone server as http://<your keystone server>:5000.
Now the scripts fails at another point, but I'll find out why and will report here.