While trying to attach a series of volumes to a set of devices, I have run into the following issue:
Some of the devices get stuck in an "infinite" state of attachment. (I do not already have volumes attached to those devices that "fail" nor are the volumes to be attached, already attached to another device)
- /dev/sde1 - EBS volume will not attach
- /dev/sde2 - OK
- /dev/sde3 - EBS volume will not attach
- /dev/sde4 - OK
- /dev/sde5 - OK
- /dev/sde6 - OK
According to the AWS documentation, these are well within the range of acceptable device mount points.
Here is some additional information about the state of the attachment.
When requesting information about the volume via the API:
ec2-describe-volumes vol-xxxxxxxx
This is the result:
ATTACHMENT vol-xxxxxxxx i-90909090 /dev/sde3 attaching 2013-05-06T21:15:56+0000 false
When you call the aws api will return an error in this case - please send it. A different approach will be to make snapshots of the original volumes, then convert it to new volumes and then attach it.
It turns out that the AWS documentation (as linked above) is somewhat ambiguous.
The documentation states that the following mount points are reserved for Instance Store Volumes:
It turns out, after a lengthy discussion with AWS support, that the following mount points are also reserved:
While the reservation does not guarantee that there will be issues, behavior by these mount points can be erratic. (As I experienced)