Ensure that many disks can be created

[waynew]

https://github.com/saltstack/salt/issues/60564

Like mentioned in that issue, when a device is created with disks we should increment. Alternatively if we really need random instead of sequential, we should just shuffle the list and pop items off instead.

[ssbn]

Also, note that this behavior affects network adapters as well. Everywhere the 'random_key' var is used in the original clouds/vmware.py source.

[jmozd]

I've been hit by this on 3006.6, too, and ou'll likely see the same problem with 3007.1, too (the code hasn't changed). It appears to be a significant issue, we're creating VMs automatically and when there are more than two disks per VM, I've been struck with this error quite a number of times in two days (and that's just test operations, where we initiate VM creation at a rate no higher than one per minute, most often much less). I've had cases where we had to re-run the creation job three times until it worked out without this error.

As in my specific case new processes are forked for each build, I resorted to to remembering the already used keys and keep retrying until a fresh one is found. I only now read above that incremental keys would be fine, too.

rand_disk_keys = []
...
    random_key = randint( -2099, -2000)
    while random_key in rand_disk_keys:
        random_key = randint( -2099, -2000)
    rand_disk_keys.append( random_key)

If the process instance would be living longer than per-job, some automatic means to clean the cache(s) (or re-init the sequence(s) ) on the start of each individual job) would need to be found.