Fail early for too long instance names #19
Conversation
resalloc_ibm_cloud/ibm_cloud_vm.py
Outdated
| @@ -339,6 +339,11 @@ def main(): | |||
| opts.instance_name = name | |||
| opts.instance = "production" if "-prod-" in name else "devel" | |||
|
|
|||
| maxlen = 56 | |||
There was a problem hiding this comment.
is it really 56? It says 63 in UI
resalloc_ibm_cloud/ibm_cloud_vm.py
Outdated
| maxlen = 56 | ||
| if len(name) > maxlen: | ||
| log.error("Name %s is too long, max %s characters", name, maxlen) | ||
| sys.exit(1) |
There was a problem hiding this comment.
Could we do something else than error? I mean something like shrinking the name?
There was a problem hiding this comment.
Thank you for looking. Could be? But it is reproducible with:
RESALLOC_NAME=s390x_us_east /var/lib/resallocserver/resalloc_provision/ibm-cloud-vm create "copr_ibm_cloud_s390x_washington_prod_13290257_20231024_201117"
...
ibm_cloud_sdk_core.api_exception.ApiException: Error: Expected only one oneOf fields to be set: got 0, Code: 400
While it is just ok with one byte dropped:
RESALLOC_NAME=s390x_us_east /var/lib/resallocserver/resalloc_provision/ibm-cloud-vm create "copr_ibm_cloud_s390x_washington_prod_13290257_20231024_20111"
...
I'd swear I was able to reproduce it with copr-ibm-cloud-s390x-washington-dev-02817070-20231024-10x string while bisecting. See differences:
copr_ibm_cloud_s390x_washington_prod_13290257_20231024_201117 fails (61 bytes)
copr_ibm_cloud_s390x_washington_prod_13290257_20231024_20111 works (60 bytes)
copr-ibm-cloud-s390x-washington-dev-02817070-20231024-10x works now, failed before
There was a problem hiding this comment.
Ok, now I think the problem was the name of the volume, which is basically instance name + suffix. PTAL
praiskup
force-pushed
the
praiskup-startup-failure
branch
from
bf551bf
to
7f10091
Compare
![github-advanced-security[bot]](https://avatars.githubusercontent.com/in/57789?s=60&v=4){kind=small}
The resource name length in IBM Cloud needs to be <= 63 characters.
If we attempt to allocate a set of resources (machine, volumes, IP, ...)
and one of the names for those resources is too long, the library fails
with pretty cryptic traceback:
Traceback (most recent call last):
File "/usr/bin/resalloc-ibm-cloud-vm", line 8, in <module>
sys.exit(main())
^^^^^^
File "/usr/lib/python3.11/site-packages/resalloc_ibm_cloud/ibm_cloud_vm.py", line 348, in main
create_instance(service, name, opts)
File "/usr/lib/python3.11/site-packages/resalloc_ibm_cloud/ibm_cloud_vm.py", line 182, in create_instance
response = service.create_instance(instance_prototype_model)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/usr/lib/python3.11/site-packages/ibm_vpc/vpc_v1.py", line 4583, in create_instance
response = self.send(request, **kwargs)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/usr/lib/python3.11/site-packages/ibm_cloud_sdk_core/base_service.py", line 341, in send
raise ApiException(response.status_code, http_response=response)
ibm_cloud_sdk_core.api_exception.ApiException: Error: Expected only one oneOf fields to be set: got 0, Code: 400
With this fix, we'll get the failure much faster:
ERROR: Field boot_volume_attachment.volume.name is longer than 63 characters: copr-ibm-cloud-s390x-washington-prod-13290257-20231024-201117-root
praiskup
force-pushed
the
praiskup-startup-failure
branch
from
7f10091
to
79e01a0
Compare
| return | ||
| log.error("Field %s is longer than %s characters: %s", | ||
| ".".join(itemspec), max_length, to_check) | ||
| sys.exit(1) |
There was a problem hiding this comment.
still wondering if failing the creation of VM and trying it again will result (after some retries) in a shorter name? Could we also reconsider (e.g. in separate PR) different and shorter naming scheme for volumes?
There was a problem hiding this comment.
still wondering if failing the creation of VM and trying it again will result (after some retries) in a shorter name?
No, the whole point of this PR is to make the failure 100% obvious.
Could we also reconsider (e.g. in separate PR) a different and shorter naming scheme for volumes?
We could. Listing them in the ibm-cloud-list-vms might be a bit more complicated then.
The whole story was basically a trivial mistake while creating the Resalloc pool name ... so I changed the
pool name from ..ibm_cloud.. to just ..ic.. and done. This PR will help others to spot the issue much easily.
The resource name length in IBM Cloud needs to be <= 63 characters.
If we attempt to allocate a set of resources (machine, volumes, IP, ...)
and one of the names for those resources is too long, the library fails
with pretty cryptic traceback:
With this fix, we'll get the failure much faster: