We are experiencing some odd behavior when using some of the new direct attached storage vms which appears to conflict somewhat with the documentation. The vms we are testing with are the i4i and the i3en instances and was expecting to simply be able to replace the ephemeral disk with the instance storage by applying the use_instance_storage flag.
We performed the following four tests and only one of the configuration appears to boot. The below is from a i3en.2xlarge which has two drives attached but we have also performed the same tests on i4i with a single disk.
It appears that the flag encrypted: null must be set or an error for incorrect mapping is thrown. Bosh always appears to create an EBS volume mounted as /dev/xvda and I thought it was failing due to the drive size of 5gb not being enough to load the bosh agent. This doesn't explain why our Test-2 fails though.
We are testing with the bosh jumpbox release and can not ssh to any of the vms other than the one which has booted to investigate further.
Bosh Director. Version 280.0.9 Director Stemcell ubuntu-22.04.1/1.301
--- Test-1
- cloud_properties:
instance_type: i3en.2xlarge
raw_instance_storage: true
/dev/xvda 5GB
/dev/sdb 10GB
Timeout on boot.
--- Test-2
- cloud_properties:
instance_type: i3en.2xlarge
raw_instance_storage: true
root_disk:
size: 30360
type: gp3
/dev/xvda 30GB
/dev/sdb 10GB
Timeout on boot.
--- Test-3
- cloud_properties:
ephemeral_disk:
encrypted: null
use_instance_storage: true
instance_type: i3en.2xlarge
root_disk:
size: 30360
type: gp3
/dev/xvda 30GB
NAME MAJ:MIN RM SIZE RO TYPE MOUNTPOINTS
nvme2n1 259:0 0 2.3T 0 disk
nvme1n1 259:1 0 2.3T 0 disk
nvme0n1 259:2 0 30G 0 disk
├─nvme0n1p1 259:3 0 4.8G 0 part /home
│ /
├─nvme0n1p2 259:4 0 12.6G 0 part [SWAP]
└─nvme0n1p3 259:5 0 12.6G 0 part /var/tmp
/tmp
/opt
/var/opt
/var/log
/var/vcap/data
--- Test-4
- cloud_properties:
ephemeral_disk:
encrypted: null
use_instance_storage: true
instance_type: i3en.2xlarge
name: i3en.2xlarge
/dev/xvda 5GB
Timeout on boot.
We have written an os-conf script to raid and mount the disks on Test-3 which is giving some very good performance results but believe the current cloud-config configuration we are using maybe incorrect or the cpi might not be working correctly.
We are experiencing some odd behavior when using some of the new direct attached storage vms which appears to conflict somewhat with the documentation. The vms we are testing with are the i4i and the i3en instances and was expecting to simply be able to replace the ephemeral disk with the instance storage by applying the use_instance_storage flag.
We performed the following four tests and only one of the configuration appears to boot. The below is from a i3en.2xlarge which has two drives attached but we have also performed the same tests on i4i with a single disk.
It appears that the flag
encrypted: nullmust be set or an error for incorrect mapping is thrown. Bosh always appears to create an EBS volume mounted as/dev/xvdaand I thought it was failing due to the drive size of 5gb not being enough to load the bosh agent. This doesn't explain why our Test-2 fails though.We are testing with the bosh jumpbox release and can not ssh to any of the vms other than the one which has booted to investigate further.
Bosh Director. Version 280.0.9 Director Stemcell ubuntu-22.04.1/1.301We have written an os-conf script to raid and mount the disks on Test-3 which is giving some very good performance results but believe the current cloud-config configuration we are using maybe incorrect or the cpi might not be working correctly.