The interesting decision in this commit was to expose AR's state and not
a fully materialized Allocation struct. AR.clientAlloc builds an Alloc
that contains the task state, so I considered simply memoizing and
exposing that method.
However, that would lead to AR having two awkwardly similar methods:
- Alloc() - which returns the server-sent alloc
- ClientAlloc() - which returns the fully materialized client alloc
Since ClientAlloc() could be memoized it would be just as cheap to call
as Alloc(), so why not replace Alloc() entirely?
Replacing Alloc() entirely would require Update() to immediately
materialize the task states on server-sent Allocs as there may have been
local task state changes since the server received an Alloc update.
This quickly becomes difficult to reason about: should Update hooks use
the TaskStates? Are state changes caused by TR Update hooks immediately
reflected in the Alloc? Should AR persist its copy of the Alloc? If so,
are its TaskStates canonical or the TaskStates on TR?
So! Forget that. Let's separate the static Allocation from the dynamic
AR & TR state!
- AR.Alloc() is for static Allocation access (often for the Job)
- AR.AllocState() is for the dynamic AR & TR runtime state (deployment
status, task states, etc).
If code needs to know the status of a task: AllocState()
If code needs to know the names of tasks: Alloc()
It should be very easy for a developer to reason about which method they
should call and what they can do with the return values.
* Stopping an alloc is implemented via Updates but update hooks are
*not* run.
* Destroying an alloc is a best effort cleanup.
* AllocRunner destroy hooks implemented.
* Disk migration and blocking on a previous allocation exiting moved to
its own package to avoid cycles. Now only depends on alloc broadcaster
instead of also using a waitch.
* AllocBroadcaster now only drops stale allocations and always keeps the
latest version.
* Made AllocDir safe for concurrent use
Lots of internal contexts that are currently unused. Unsure if they
should be used or removed.
Tons left to do and lots of churn:
1. No state saving
2. No shutdown or gc
3. Removed AR factory *for now*
4. Made all "Config" structs local to the package they configure
5. Added allocID to GC to avoid a lookup
Really hating how many things use *structs.Allocation. It's not bad
without state saving, but if AllocRunner starts updating its copy things
get racy fast.
This PR does:
1. Health message based on detection has format "Driver XXX detected"
and "Driver XXX not detected"
2. Set initial health description based on detection status and don't
wait for the first health check.
3. Combine updating attributes on the node, fingerprint and health
checking update for drivers into a single call back.
4. Condensed driver info in `node status` only shows detected drivers
and make the output less wide by removing spaces.
Handle the ErrNoLeader case and apply slower retries. Also when we have
missed the heartbeat retry aggressively, backing off after we have
missed for more than 30 seconds.
Fix a bug in which if the node attributes/meta changed, we would
re-register the node in status initializing. This would incorrectly
trigger the client to log that it missed its heartbeat.
It would change the status of the Node to initializing until the next
heartbeat occured.
* Allow server TLS configuration to be reloaded via SIGHUP
* dynamic tls reloading for nomad agents
* code cleanup and refactoring
* ensure keyloader is initialized, add comments
* allow downgrading from TLS
* initalize keyloader if necessary
* integration test for tls reload
* fix up test to assert success on reloaded TLS configuration
* failure in loading a new TLS config should remain at current
Reload only the config if agent is already using TLS
* reload agent configuration before specific server/client
lock keyloader before loading/caching a new certificate
* introduce a get-or-set method for keyloader
* fixups from code review
* fix up linting errors
* fixups from code review
* add lock for config updates; improve copy of tls config
* GetCertificate only reloads certificates dynamically for the server
* config updates/copies should be on agent
* improve http integration test
* simplify agent reloading storing a local copy of config
* reuse the same keyloader when reloading
* Test that server and client get reloaded but keep keyloader
* Keyloader exposes GetClientCertificate as well for outgoing connections
* Fix spelling
* correct changelog style
The Client.allocs map now contains all AllocRunners again, not just
un-GC'd AllocRunners. Client.allocs is only pruned when the server GCs
allocs.
Also stops logging "marked for GC" twice.
Fixes https://github.com/hashicorp/nomad/issues/3454
Reliably reproduced the data race before by having a fingerprinter
change the nodes attributes every millisecond and syncing at the same
rate. With fix, did not ever panic.
This PR removes locking around commonly accessed node attributes that do
not need to be locked. The locking could cause nodes to TTL as the
heartbeat code path was acquiring a lock that could be held for an
excessively long time. An example of this is when Vault is inaccessible,
since the fingerprint is run with a lock held but the Vault
fingerprinter makes the API calls with a large timeout.
Fixes https://github.com/hashicorp/nomad/issues/2689
interface has 3 implementations:
1. local for blocking and moving data locally
2. remote for blocking and moving data from another node
3. noop for allocs that don't need to block
Fixes#2478#2474#1995#2294
The new client only handles agent and task service advertisement. Server
discovery is mostly unchanged.
The Nomad client agent now handles all Consul operations instead of the
executor handling task related operations. When upgrading from an
earlier version of Nomad existing executors will be told to deregister
from Consul so that the Nomad agent can re-register the task's services
and checks.
Drivers - other than qemu - now support an Exec method for executing
abritrary commands in a task's environment. This is used to implement
script checks.
Interfaces are used extensively to avoid interacting with Consul in
tests that don't assert any Consul related behavior.
This PR takes the host ID and runs it through a hash so that it is well
distributed. This makes it so that machines that report similar host IDs
are easily distinguished.
Instances of similar IDs occur on EC2 where the ID is prefixed and on
motherboards created in the same batch.
Fixes https://github.com/hashicorp/nomad/issues/2534
This PR adds tracking to when a task starts and finishes and the logs
API takes advantage of this and returns better errors when asking for
logs that do not exist.
This PR fixes two issues:
* Folder permissions in -dev mode were incorrect and not suitable for
running as a particular user.
* Was not setting the group membership properly for the launched
process.
Fixes https://github.com/hashicorp/nomad/issues/2160
This PR introduces a parallelism limit during garbage collection. This
is used to avoid large resource usage spikes if garbage collecting many
allocations at once.
This PR adds the following metrics to the client:
client.allocations.migrating
client.allocations.blocked
client.allocations.pending
client.allocations.running
client.allocations.terminal
Also adds some missing fields to the API version of the evaluation.
This PR fixes our vet script and fixes all the missed vet changes.
It also fixes pointers being printed in `nomad stop <job>` and `nomad
node-status <node>`.
This PR makes GetAllocs use a blocking query as well as adding a sanity
check to the clients watchAllocation code to ensure it gets the correct
allocations.
This PR fixes https://github.com/hashicorp/nomad/issues/2119 and
https://github.com/hashicorp/nomad/issues/2153.
The issue was that the client was talking to two different servers, one
to check which allocations to pull and the other to pull those
allocations. However the latter call was not with a blocking query and
thus the client would not retreive the allocations it requested.
The logging has been improved to make the problem more clear as well.
Client.Shutdown holds the allocLock when destroying alloc runners in dev
mode.
Client.updateAllocStatus can be called during AllocRunner shutdown and
calls getAllocRunners which tries to acquire allocLock.RLock. This
deadlocks since Client.Shutdown already has the write lock.
Switching Client.Shutdown to use getAllocRunners and not hold a lock
during AllocRunner shutdown is the solution.
The Driver.Prestart method currently does very little but lays the
foundation for where lifecycle plugins can interleave execution _after_
task environment setup but _before_ the task starts.
Currently Prestart does two things:
* Any driver specific task environment building
* Download Docker images
This change also attaches a TaskEvent emitter to Drivers, so they can
emit events during task initialization.
This PR disallows stale queries when deriving a Vault token. Allowing
stale queries could result in the allocation not existing on the server
that is servicing the request.
