-
Notifications
You must be signed in to change notification settings - Fork 930
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Port virtualbox scripts to VBoxManage CLI #625
base: main
Are you sure you want to change the base?
Conversation
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Thanks for the work @stevemk14ebr! I need to still test the code locally, but I have added some questions and improvement suggestions already. It is good to see what we can do with VBoxManage and how it allows us to remove the virtualbox dependency. The disadvantage is that it is less flexible, as it does not export everything in the API (for example, it seems it is not possible to access the max number of adapters which would allow us to write simpler code as in the previous version) and that we need to create a subprocess everytime we want to run a command. The new code using VBoxManage also looks longer and more complicated, but we may be able to simplify it a bit.
What about keeping both the version using the virtualbox library and the new one using VBoxManage until we have tested and migrated everything else?
Also, I think we need some documentation in /virtualbox/README.md.
virtualbox/vbox-export-snapshots.py
Outdated
except subprocess.CalledProcessError as e: | ||
# exit code is an error | ||
print(f"Error running VBoxManage command: {e} ({e.stderr})") | ||
raise Exception(f"Error running VBoxManage command") |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
why is it needed to catch the exception to print and error and re-reise it? I see the same pattern in other functions as well.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Style choice, this throws a pretty error to the top level main to print out. I can change if you think there's a more pythonic style
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
I have seen in other Python code that an exception is re-triggered to add extra details or format the exception differently, but without a print that duplicates similar information. The print apart from duplicating the information, can make the output difficult to digest in this case, as {e.stderr}
is rendering the output that is likely to be the long help message from VBoxManage. I think we should remove the try-catch, as the re-triggered exception is almost the same:
- Original exception:
Command '['VBoxManage', 'list2', 'list', 'hostonlyifs']' returned non-zero exit status 2
- Retriggered exception:
Error running VBoxManage command: Command '['VBoxManage', 'list2', 'list', 'hostonlyifs']' returned non-zero exit status 2.
except subprocess.CalledProcessError as e: | |
# exit code is an error | |
print(f"Error running VBoxManage command: {e} ({e.stderr})") | |
raise Exception(f"Error running VBoxManage command") |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Also, if we re-raise the exception, I think we should use a more concrete typ of exception like RuntimeError
.
session.unlock_machine() | ||
print(f"Restored '{snapshot_name}' and changed its adapter(s) to host-only") | ||
|
||
vm_uuid = get_vm_uuid(VM_NAME) |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
do we need to get the UUID? It seems like the commands work with the VM_NAME (we may need to enclose the entire name in double quotes to avoid issues with spaces), or am I missing something?
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
We could rely on the VM_NAME alone, but I use the UUID so that we can support multiple VMs of the same name and be sure we refer to the same VM consistently for all operations
we can, the vminfo command lists all 8 adapters (the max) and any unset adapters have the value 'none'. The code doesn't need to check the max adapters because it lists all of them, even if they're unset, so we always loop all 8 adapters.
I have no issues with not merging these PRs (I will send more for the other two scripts) until we are ready to drop the |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
I have tested vbox-export-snapshots.py
and I it fails because the network interface does not have a name. I though this happened in the previous version when exporting the VM, but it seems like setting it could be the issue and because you start the VM (what I was not doing in the previous version) it fails even before exporting it:
Starting operations on FLARE-VM
VM {40138663-f254-412b-8776-10a7cc08daea} is already shut down (state: poweroff).
Restored 'FLARE-VM'
VM {40138663-f254-412b-8776-10a7cc08daea} is already shut down (state: poweroff).
Found existing hostonlyif vboxnet0
Changed nic1
Nic configuration verified correct
Power cycling before export...
VM {40138663-f254-412b-8776-10a7cc08daea} is not running (state: poweroff). Starting VM...
Error running VBoxManage command: Command '['VBoxManage', 'startvm', '{40138663-f254-412b-8776-10a7cc08daea}', '--type', 'gui']' returned non-zero exit status 1. (VBoxManage: error: Nonexistent host networking interface, name '' (VERR_INTERNAL_ERROR)
VBoxManage: error: Details: code NS_ERROR_FAILURE (0x80004005), component ConsoleWrap, interface IConsole
)
Error checking VM state: Error running VBoxManage command
Unexpectedly failed doing operations on FLARE-VM. Exiting...
Done. Exiting...
I reported what I think was a bug in https://www.virtualbox.org/ticket/22158. But what really confuses me is that it seems it does work for you. 😕
virtualbox/vbox-export-snapshots.py
Outdated
except subprocess.CalledProcessError as e: | ||
# exit code is an error | ||
print(f"Error running VBoxManage command: {e} ({e.stderr})") | ||
raise Exception(f"Error running VBoxManage command") |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
I have seen in other Python code that an exception is re-triggered to add extra details or format the exception differently, but without a print that duplicates similar information. The print apart from duplicating the information, can make the output difficult to digest in this case, as {e.stderr}
is rendering the output that is likely to be the long help message from VBoxManage. I think we should remove the try-catch, as the re-triggered exception is almost the same:
- Original exception:
Command '['VBoxManage', 'list2', 'list', 'hostonlyifs']' returned non-zero exit status 2
- Retriggered exception:
Error running VBoxManage command: Command '['VBoxManage', 'list2', 'list', 'hostonlyifs']' returned non-zero exit status 2.
except subprocess.CalledProcessError as e: | |
# exit code is an error | |
print(f"Error running VBoxManage command: {e} ({e.stderr})") | |
raise Exception(f"Error running VBoxManage command") |
virtualbox/vbox-export-snapshots.py
Outdated
except subprocess.CalledProcessError as e: | ||
# exit code is an error | ||
print(f"Error running VBoxManage command: {e} ({e.stderr})") | ||
raise Exception(f"Error running VBoxManage command") |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Also, if we re-raise the exception, I think we should use a more concrete typ of exception like RuntimeError
.
Co-authored-by: Ana María Martínez Gómez <[email protected]>
I did some more testing. Exporting a VM setting the hostonly adapter (with either virtualbox API and VBoxManage CLI) fails when I have never used the hostonly adapter of the VM as exporting does not set the name. It does not appear possible to use the API/VBoxManage CLI and I still think this is a virtualbox bug as reported on as reported on https://www.virtualbox.org/ticket/22158. But for our case where we always use the same VM to export several snapshots, we can ensure the hostonly adapter has a name before creating the snapshots: Set the network to hostonly (save the settings) and then back to NAT (save setting again). This ensures the hostonly adapter name is set and then the exporting using the virtualbox API and VBoxManage CLI works. So the issue is not a blocker for this PR. Thanks @stevemk14ebr for working on this! This is a very intuitive bug and your work was very helpful to figure out a fix. 💐 |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
this file should not be commited.
import sys | ||
import textwrap | ||
import argparse | ||
import virtualbox | ||
from virtualbox.library import NetworkAttachmentType as NetType | ||
import subprocess | ||
import re | ||
import time | ||
import gi | ||
gi.require_version('Notify', '0.7') | ||
from gi.repository import Notify | ||
from vboxcommon import * |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
recommend using isort for formatting consistency
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
and black for other formatting consistency
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
not blockers, but might be nice
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Good idea @williballenthin! We can address this after this PR has been merged in a different PR, added to #507 (comment). So that @stevemk14ebr does not need to add more things to this PR. 😉
DISABLED_ADAPTER_TYPE = "hostonly" | ||
ALLOWED_ADAPTER_TYPES = ("hostonly", "intnet", "none") | ||
|
||
def get_vm_uuids(dynamic_only): |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
recommend type hints, at least for function signatures, as a form of documentation
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
not a blocker, but nice to have
if dynamic_only and DYNAMIC_VM_NAME in vm_name: | ||
machine_guids.append((vm_name, machine_guid)) | ||
else: | ||
machine_guids.append((vm_name, machine_guid)) |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
i dont understand this logic, aren't the branches the same on both sides?
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
This change was made to try to address my feedback from #625 (comment). We want to add all names if dynamic_only
is not set, and only the VMs with .dynamic
in the name if it is set. I think it should be:
if dynamic_only and DYNAMIC_VM_NAME in vm_name: | |
machine_guids.append((vm_name, machine_guid)) | |
else: | |
machine_guids.append((vm_name, machine_guid)) | |
if (not dynamic_only) or DYNAMIC_VM_NAME in vm_name: | |
machine_guids.append((vm_name, machine_guid)) |
machine_guids = [] | ||
try: | ||
vms_output = run_vboxmanage(["list", "vms"]) | ||
pattern = r'"(.*?)" \{(.*?)\}' |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
when dealing with regular expressions (and command output), i'd recommend including a few example lines of the pattern text, so its easy for a human to follow along. otherwise, i have to guess what this pattern does, which can be hard or impossible for some regular expressions.
Args: | ||
vm_name: The name of the VM. | ||
snapshot_name: The name of the snapshot. | ||
|
||
Returns: | ||
A list of snapshot names that are children of the given snapshot. |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
nice documentation!
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
does this return all descendents recursively? or only the direct children?
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
It should return all. I think it would be good to clarify it in the documentation. 👍
snapshot_regex = rf'(SnapshotName(?:-\d+)*)=\"(.*?)\"' | ||
snapshots = re.findall(snapshot_regex, vminfo, flags=re.M) | ||
|
||
children = [] | ||
|
||
# find the root SnapshotName by matching the name | ||
root_snapshotid = None | ||
for snapshotid, snapshot_name in snapshots: | ||
if snapshot_name.lower() == root_snapshot_name.lower() and (not any(p.lower() in snapshot_name.lower() for p in protected_snapshots)): | ||
root_snapshotid = snapshotid | ||
|
||
if not root_snapshotid: | ||
print("Failed to find root snapshot") | ||
raise Exception(f"Failed to find root snapshot {snapshot_name}") | ||
|
||
# children of that snapshot share the same prefix id | ||
dependant_child = False | ||
for snapshotid, snapshot_name in snapshots: | ||
if snapshotid.startswith(root_snapshotid): | ||
if not any(p.lower() in snapshot_name.lower() for p in protected_snapshots): | ||
children.append((snapshotid, snapshot_name)) | ||
else: | ||
dependant_child = True | ||
|
||
# remove the root snapshot if any children are protected OR it's the current snapshot | ||
if dependant_child: | ||
print("Root snapshot cannot be deleted as a child snapshot is protected") | ||
children = [snapshot for snapshot in children if snapshot[0] != root_snapshotid] |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
i think i can follow this logic, and it seems reasonable, but without some example output, i really have any idea.
furthermore, i wouldn't be comfortable changing this logic without any tests. how would i know if i broke anything (especially without vboxmanage installed)?
i'd recommend making this parsing region its own function (input: str, output: list), and then adding some test cases with known data. the test cases will serve both as documentation (since readers can see some examples of real-world data) and to prevent regressions.
vm = vbox.find_machine(vm_name) | ||
snapshot = vm.find_snapshot(snapshot_name) | ||
get_snapshots_to_delete(snapshot, protected_snapshots) | ||
TO_DELETE = get_snapshot_children(vm_name, snapshot_name, protected_snapshots) |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
consider reserving upper snake case names for constants
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
not a blocker
print(f"\nVM state: {vm.state}\n⚠️ Snapshot deleting is slower in a running VM and may fail in a changing state") | ||
vm_state = get_vm_state(vm_name) | ||
if vm_state not in ("poweroff", "saved"): | ||
print(f"\nVM state: {vm_state}\n⚠️ Snapshot deleting is slower in a running VM and may fail in a changing state") | ||
|
||
answer = input("\nConfirm deletion ('y'):") |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
is "y" the default, or is that what you should press to confirm?
from the code i see its what you should press, but i'm afraid that in this format the message suggests "y" is the default. maybe change to "press 'y'".
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
The idea is that the user has to press 'y', as this deletes the snapshots and there is no way to recover them. I agree it would be good to clarify it.
|
||
answer = input("\nConfirm deletion ('y'):") | ||
if answer.lower() == "y": | ||
print("\nDeleting... (this may take some time, go for an 🍦!)") | ||
session = vm.create_session() | ||
for name, uuid in TO_DELETE: | ||
for snapshotid, snapshot_name in TO_DELETE[::-1]: # delete in reverse order to avoid issues with child snapshots |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
for snapshotid, snapshot_name in TO_DELETE[::-1]: # delete in reverse order to avoid issues with child snapshots | |
for snapshotid, snapshot_name in reversed(TO_DELETE): # delete in reverse order to avoid issues with child snapshots |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
also, interesting behavior. i think get_snapshot_children
should document that the order of the result has a particular meaning.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
This was a bug in my original script, that @stevemk14ebr corrected here. I think it is a good idea to document this in the get_snapshot_children
function. 👍
|
Ports to VBoxManage CLI, identical logic otherwise. Errors handled gracefully for the most part. Output: