This pull request was automerged via backport-assistant

Merge pull request #12605 from hashicorp/backport/fix/multi-task-group-canary-deploys/sincerely-subtle-husky

This pull request was automerged via backport-assistant

Co-authored-by: Sarah <sthompson@hashicorp.com>
Co-authored-by: Luiz Aoqui <luiz@hashicorp.com>

The spread iterator can panic when processing an evaluation, resulting
in an unrecoverable state in the cluster. Whenever a panicked server
restarts and quorum is restored, the next server to dequeue the
evaluation will panic.

To trigger this state:
* The job must have `max_parallel = 0` and a `canary >= 1`.
* The job must not have a `spread` block.
* The job must have a previous version.
* The previous version must have a `spread` block and at least one
  failed allocation.

In this scenario, the desired changes include `(place 1+) (stop
1+), (ignore n) (canary 1)`. Before the scheduler can place the canary
allocation, it tries to find out which allocations can be
stopped. This passes back through the stack so that we can determine
previous-node penalties, etc. We call `SetJob` on the stack with the
previous version of the job, which will include assessing the `spread`
block (even though the results are unused). The task group spread info
state from that pass through the spread iterator is not reset when we
call `SetJob` again. When the new job version iterates over the
`groupPropertySets`, it will get an empty `spreadAttributeMap`,
resulting in an unexpected nil pointer dereference.

This changeset resets the spread iterator internal state when setting
the job, logging with a bypass around the bug in case we hit similar
cases, and a test that panics the scheduler without the patch.

Add new namespace ACL requirement for the /v1/jobs/parse endpoint and
return early if HCLv2 parsing fails.

The endpoint now requires the new `parse-job` ACL capability or
`submit-job`.

This PR adds symlink resolution when doing validation of paths
to ensure they do not escape client allocation directories.

go-getter creates a circular dependency between a Client and Getter,
which means each is inherently thread-unsafe if you try to re-use
on or the other.

This PR fixes Nomad to no longer make use of the default Getter objects
provided by the go-getter package. Nomad must create a new Client object
on every artifact download, as the Client object controls the Src and Dst
among other things. When Caling Client.Get, the Getter modifies its own
Client reference, creating the circular reference and race condition.

We can still achieve most of the desired connection caching behavior by
re-using a shared HTTP client with transport pooling enabled.

When an allocation stops, the `csi_hook` makes an unpublish RPC to the
servers to unpublish via the CSI RPCs: first to the node plugins and
then the controller plugins. The controller RPCs must happen after the
node RPCs so that the node has had a chance to unmount the volume
before the controller tries to detach the associated device.

But the client has local access to the node plugins and can
independently determine if it's safe to send unpublish RPC to those
plugins. This will allow the server to treat the node plugin as
abandoned if a client is disconnected and `stop_on_client_disconnect`
is set. This will let the server try to send unpublish RPCs to the
controller plugins, under the assumption that the client will be
trying to unmount the volume on its end first.

Note that the CSI `NodeUnpublishVolume`/`NodeUnstageVolume` RPCs can 
return ignorable errors in the case where the volume has already been
unmounted from the node. Handle all other errors by retrying until we
get success so as to give operators the opportunity to reschedule a
failed node plugin (ex. in the case where they accidentally drained a
node without `-ignore-system`). Fan-out the work for each volume into
its own goroutine so that we can release a subset of volumes if only
one is stuck.

Small refactoring of the allocrunner hook for CSI to make it more
testable, and a unit test that covers most of its logic.

* The volume claim GC method and volumewatcher both have logic
collecting terminal allocations that duplicates most of the logic
that's now in the state store's `CSIVolumeDenormalize` method. Copy
this logic into the state store so that all code paths have the same
view of the past claims.
* Remove logic in the volume claim GC that now lives in the state
store's `CSIVolumeDenormalize` method.
* Remove logic in the volumewatcher that now lives in the state
store's `CSIVolumeDenormalize` method.
* Remove logic in the node unpublish RPC that now lives in the state
store's `CSIVolumeDenormalize` method.

In the client's `(*csiHook) Postrun()` method, we make an unpublish
RPC that includes a claim in the `CSIVolumeClaimStateUnpublishing`
state and using the mode from the client. But then in the
`(*CSIVolume) Unpublish` RPC handler, we query the volume from the
state store (because we only get an ID from the client). And when we
make the client RPC for the node unpublish step, we use the _current
volume's_ view of the mode. If the volume's mode has been changed
before the old allocations can have their claims released, then we end
up making a CSI RPC that will never succeed.

Why does this code path get the mode from the volume and not the
claim? Because the claim written by the GC job in `(*CoreScheduler)
csiVolumeClaimGC` doesn't have a mode. Instead it just writes a claim
in the unpublishing state to ensure the volumewatcher detects a "past
claim" change and reaps all the claims on the volumes.

Fix this by ensuring that the `CSIVolumeDenormalize` creates past
claims for all nil allocations with a correct access mode set.

* csi: resolve invalid claim states on read

It's currently possible for CSI volumes to be claimed by allocations
that no longer exist. This changeset asserts a reasonable state at
the state store level by registering these nil allocations as "past
claims" on any read. This will cause any pass through the periodic GC
or volumewatcher to trigger the unpublishing workflow for those claims.

* csi: make feasibility check errors more understandable

When the feasibility checker finds we have no free write claims, it
checks to see if any of those claims are for the job we're currently
scheduling (so that earlier versions of a job can't block claims for
new versions) and reports a conflict if the volume can't be scheduled
so that the user can fix their claims. But when the checker hits a
claim that has a GCd allocation, the state is recoverable by the
server once claim reaping completes and no user intervention is
required; the blocked eval should complete. Differentiate the
scheduler error produced by these two conditions.

The volumewatcher that runs on the leader needs to make RPC calls
rather than writing to raft (as we do in the deploymentwatcher)
because the unpublish workflow needs to make RPC calls to the
clients. This requires that the volumewatcher has access to the
leader's ACL token.

But when leadership transitions, the new leader creates a new leader
ACL token. This ACL token needs to be passed into the volumewatcher
when we enable it, otherwise the volumewatcher can find itself with a
stale token.

When `volumewatcher.Watcher` starts on the leader, it starts a watch
on every volume and triggers a reap of unused claims on any change to
that volume. But if a reaping is in-flight during leadership
transitions, it will fail and the event that triggered the reap will
be dropped. Perform one reap of unused claims at the start of the
watcher so that leadership transitions don't drop this event.

Fix Node.Copy()

When we copy the system DNS to a task's `resolv.conf`, we should set
the permissions as world-readable so that unprivileged users within
the task can read it.

The size of `stat_t` fields is architecture dependent, which was
reportedly causing a build failure on FreeBSD ARM7 32-bit
systems. This changeset matches the behavior we have on Linux.

When the `volume deregister` or `volume detach` commands get an ID
prefix that matches multiple volumes, show the full length of the
volume IDs in the list of volumes shown so so that the user can select
the correct one.

The command line client sends a specific volume ID, but this isn't
enforced at the API level and we were incorrectly using a prefix match
for volume deregistration, resulting in cases where a volume with a
shorter ID that's a prefix of another volume would be deregistered
instead of the intended volume.

changelog: add entry for #11848