We use capped exponential backoff in several places in the code when handling
failures. The code we've copy-and-pasted all over has a check to see if the
backoff is greater than the limit, but this check happens after the bitshift and
we always increment the number of attempts. This causes an overflow with a
fairly small number of failures (ex. at one place I tested it occurs after only
24 iterations), resulting in a negative backoff which then never recovers. The
backoff becomes a tight loop consuming resources and/or DoS'ing a Nomad RPC
handler or an external API such as Vault. Note this doesn't occur in places
where we cap the number of iterations so the loop breaks (usually to return an
error), so long as the number of iterations is reasonable.
Introduce a helper with a check on the cap before the bitshift to avoid overflow in all
places this can occur.
Fixes: #18199
Co-authored-by: stswidwinski <stan.swidwinski@gmail.com>
Upon dequeuing an evaluation workers snapshot their state store at the
eval's wait index or later. This ensures we process an eval at a point
in time after it was created or updated. Processing an eval on an old
snapshot could cause any number of problems such as:
1. Since job registration atomically updates an eval and job in a single
raft entry, scheduling against indexes before that may not have the
eval's job or may have an older version.
2. The older the scheduler's snapshot, the higher the likelihood
something has changed in the cluster state which will cause the plan
applier to reject the scheduler's plan. This could waste work or
even cause eval's to be failed needlessly.
However, the workers run in parallel with a new server pulling the
cluster state from a peer. During this time, which may be many minutes
long, the state store is likely far behind the minimum index required
to process evaluations.
This PR addresses this by adding an additional long backoff period after
an eval is nacked. If the scheduler's indexes catches up within the
additional backoff, it will unblock early to dequeue the next eval.
When the server shuts down we'll get a `context.Canceled` error from the state
store method. We need to bubble this error up so that other callers can detect
it. Handle this case separately when waiting after dequeue so that we can warn
on shutdown instead of throwing an ambiguous error message with just the text
"canceled."
While there may be more precise ways to block scheduling until the
server catches up, this approach adds little risk and covers additional
cases where a server may be temporarily behind due to a spike in load or
a saturated network.
For testing, we make the `raftSyncLimit` into a parameter on the worker's `run` method
so that we can run backoff tests without waiting 30+ seconds. We haven't followed thru
and made all the worker globals into worker parameters, because there isn't much
use outside of testing, but we can consider that in the future.
Co-authored-by: Tim Gross <tgross@hashicorp.com>
* One-time tokens are not replicated between regions, so we don't want to enforce
that the version check across all of serf, just members in the same region.
* Scheduler: Disconnected clients handling is specific to a single region, so we
don't want to enforce that the version check across all of serf, just members in
the same region.
* Variables: enforce version check in Apply RPC
* Cleans up a bunch of legacy checks.
This changeset is specific to 1.4.x and the changes for previous versions of
Nomad will be manually backported in a separate PR.
* planner: expose ServerMeetsMinimumVersion via Planner interface
* filterByTainted: add flag indicating disconnect support
* allocReconciler: accept and pass disconnect support flag
* tests: update dependent tests
## Development Environment Changes
* Added stringer to build deps
## New HTTP APIs
* Added scheduler worker config API
* Added scheduler worker info API
## New Internals
* (Scheduler)Worker API refactor—Start(), Stop(), Pause(), Resume()
* Update shutdown to use context
* Add mutex for contended server data
- `workerLock` for the `workers` slice
- `workerConfigLock` for the `Server.Config.NumSchedulers` and
`Server.Config.EnabledSchedulers` values
## Other
* Adding docs for scheduler worker api
* Add changelog message
Co-authored-by: Derek Strickland <1111455+DerekStrickland@users.noreply.github.com>
Rename SnapshotAfter to SnapshotMinIndex. The old name was not
technically accurate. SnapshotAtOrAfter is more accurate, but wordy and
still lacks context about what precisely it is at or after (the index).
SnapshotMinIndex was chosen as it describes the action (snapshot), a
constraint (minimum), and the object of the constraint (index).
Plan application should use a state snapshot at or after the Raft index
at which the plan was created otherwise it risks being rejected based on
stale data.
This commit adds a Plan.SnapshotIndex which is set by workers when
submitting plan. SnapshotIndex is set to the Raft index of the snapshot
the worker used to generate the plan.
Plan.SnapshotIndex plays a similar role to PlanResult.RefreshIndex.
While RefreshIndex informs workers their StateStore is behind the
leader's, SnapshotIndex is a way to prevent the leader from using a
StateStore behind the worker's.
Plan.SnapshotIndex should be considered the *lower bound* index for
consistently handling plan application.
Plans must also be committed serially, so Plan N+1 should use a state
snapshot containing Plan N. This is guaranteed for plans *after* the
first plan after a leader election.
The Raft barrier on leader election ensures the leader's statestore has
caught up to the log index at which it was elected. This guarantees its
StateStore is at an index > lastPlanIndex.
The old logic for cancelling duplicate blocked evaluations by job id had
the issue where the newer evaluation could have additional node classes
that it is (in)eligible for that we would not capture. This could make
it such that cluster state could change such that the job would make
progress but no evaluation was unblocked.
Denormalize jobs in AppendAllocs:
AppendAlloc was originally only ever called for inplace upgrades and new
allocations. Both these code paths would remove the job from the
allocation. Now we use this to also add fields such as FollowupEvalID
which did not normalize the job. This is only a performance enhancement.
Ignore terminal allocs:
Failed allocations are annotated with the followup Eval ID when one is
created to replace the failed allocation. However, in the plan applier,
when we check if allocations fit, these terminal allocations were not
filtered. This could result in the plan being rejected if the node would
be overcommited if the terminal allocations resources were considered.
This PR fixes a scheduling race condition in which the plan results from
one invocation of the scheduler were not being considered by the next
since the Worker was not waiting for the correct index.
Fixes https://github.com/hashicorp/nomad/issues/3198
Luiz Aoqui