GitLab

* Add a new class for snf-manage *-remove commands

* Add assertSuccess201 function
* Use this new function in test/subnets.py

* Fix a potential crash if the gateway IP was the last available IP of
the subnet
* Create the corresponding test

* Fix a bug where snf-manage subnet-list printed deleted subnets too
* Fix a query, that also returned deleted subnets

* Allow the creation of a network without specifying any subnets

Fix handling of reboot type option. If 'hard' reboot type is specified
this is mapped to zero shutdown timeout of OP_INSTANCE_REBOOT Ganeti
job. This is only supported from Ganeti>=2.10 and snf-ganeti. In other
versions this options will be ignored, and we will fall back to Ganeti's
default shutdown timeout (120s).

Add missing 'shutdown_timeout' option to reboot, shutdown and destroy
methods of Ganeti RAPI client. The Ganeti RAPI daemon will ignore this
options until Ganeti 2.10.

Consider as default owner for a port the owner of the machine that the
port will be connected to.

Until now, NICs that did not exist in the Ganeti backend and were in
BUILD state were removed after a timeout (BUILDING_NIC_TIMEOUT). This
was dangerous, because if  an OP_INSTANCE_CONNECT job had been "stuck"
in Ganeti, the NIC and it's IP address would be released. This commit
fixes this by removing such NICs only if the OP_INSTANCE_CONNECT has
been finished and the NIC does not exist in the Ganeti backend. Because,
querying Ganeti has a time overhead, this is only done for NICs that
have not been created too recently.

Add new action 'set_user_quota' to set base quota for existing users.

Update Changelog and upgrade notes.

When re-registering an existing resource, do not override the visibility
flags. They may have been modified by the administrator.

Also fix field naming in resource-list.

Migration 0031 accepted users to whom activation email had been sent, even
if their email had not been verified. This is an inconsistent state. We
set moderated = False for inactive users. If there are any active users in
this state, we issue a warning but leave them untouched.

Restructure code to avoid an extra join and dict manipulations.

When registering a new resource or updating base quota for a single
resource, compute and set holdings just for this resource.

Introduce command enforce-resources-cyclades, which checks for quota
violations and applies actions to enforce the limits. By default, it
operates only on these resources: cpu, ram, floating_ip.

Add option `--all' in command user-modify to support setting base quota
for all users. Exceptions can be specified with `--exclude'.

Also select user for update in user-modify.

Enable filtering quota by resource, over limit status, usage and
limit values.

Introduce quota-list and quota-verify.

Make available in the API only resources with api_visible=True.

Rename flag `allow_in_projects' to `ui_visible'. Add flag `api_visible'.
The former entails the latter.

View the resource defaults as a skeleton to be consulted when accepting
a new user. All users keep their quota in AstakosUserQuota.

Operate resource-modify in bulk, in order to avoiding updating the
quotaholder separately for each resource.

Use a more descriptive name for this resource-modify option. Also change
resource-list column `limit' to `default_quota'.

Register new resources with unlimited quota.

By convention, infinite is represented by 2**63-1, the max value for
a signed 64 bit field.

Any values that happen to be greater than the new max (2**63-1)
are reduced to this value.

Avoid stateful update of holdings when a resource limit changes; rather
recompute and set all affected users' quota.

Instead of updating every single holding, delete all after loading them,
then reinsert them updated with bulk_create.

Also, remove unneeded param of astakos_users_quotas.

They have been subsumed by the respective service import/export commands.