Fixes#2093
Enable configuring `memory_hard_limit` in the docker config stanza for tasks.
If set, this field will be passed to the container runtime as `--memory`, and
the `memory` configuration from the task resource configuration will be passed
as `--memory_reservation`, creating hard and soft memory limits for tasks using
the docker task driver.
Fix some docker test flakiness where image cleanup process may
contaminate other tests. A clean up process may attempt to delete an
image while it's used by another test.
This fixes few cases where driver eventor goroutines are leaked during
normal operations, but especially so in tests.
This change makes few modifications:
First, it switches drivers to use `Context`s to manage shutdown events.
Previously, it relied on callers invoking `.Shutdown()` function that is
specific to internal drivers only and require casting. Using `Contexts`
provide a consistent idiomatic way to manage lifecycle for both internal
and external drivers.
Also, I discovered few places where we don't clean up a temporary driver
instance in the plugin catalog code, where we dispense a driver to
inspect and validate the schema config without properly cleaning it up.
When an allocation runs for a task driver that can't support volume mounts,
the mounting will fail in a way that can be hard to understand. With host
volumes this usually means failing silently, whereas with CSI the operator
gets inscrutable internals exposed in the `nomad alloc status`.
This changeset adds a MountConfig field to the task driver Capabilities
response. We validate this when the `csi_hook` or `volume_hook` fires and
return a user-friendly error.
Note that we don't currently have a way to get driver capabilities up to the
server, except through attributes. Validating this when the user initially
submits the jobspec would be even better than what we're doing here (and could
be useful for all our other capabilities), but that's out of scope for this
changeset.
Also note that the MountConfig enum starts with "supports all" in order to
support community plugins in a backwards compatible way, rather than cutting
them off from volume mounting unexpectedly.
This fixes a bug where docker images may not be GCed. The cause of the
bug is that we track the task using `task.ID+task.Name` on task start
but remove on plain `task.ID`.
This haromize the two paths by using `task.ID`, as it's unique enough
and it's also used in the `loadImage` path (path when loading an image
from a local tarball instead of dockerhub).
Makes it possible to run Linux Containers On Windows with Nomad alongside Windows Containers. Fingerprint prevents only to run Nomad in Windows 10 with Linux Containers
Protect against a panic when we attempt to start a container with a name
that conflicts with an existing one. If the existing one is being
deleted while nomad first attempts to create the container, the
createContainer will fail with `container already exists`, but we get
nil container reference from the `containerByName` lookup, and cause a
crash.
I'm not certain how we get into the state, except for being very
unlucky. I suspect that this case may be the result of a concurrent
restart or the docker engine API not being fully consistent (e.g. an
earlier call purged the container, but docker didn't free up resources
yet to create a new container with the same name immediately yet).
If that's the case, then re-attempting creation will hopefully succeed,
or we'd at least fail enough times for the alloc to be rescheduled to
another node.
Looks like the latest `github.com/docker/docker/registry.ResolveAuthConfig` expect
`github.com/docker/docker/api/types.AuthConfig` rather than
`github.com/docker/cli/cli/config/types.AuthConfig`. The two types are
identical but live in different packages.
Here, we embed `registry.ResolveAuthConfig` from upstream repo, but with
the signature we need.
This fixes a bug where executor based drivers emit stats every second,
regardless of user configuration.
When serializing the Stats request across grpc, the nomad agent dropped
the Interval value, and then executor uses 1s as a default value.
* Making pull activity timeout configurable in Docker plugin config, first pass
* Fixing broken function call
* Fixing broken tests
* Fixing linter suggestion
* Adding documentation on new parameter in Docker plugin config
* Adding unit test
* Setting min value for pull_activity_timeout, making pull activity duration a private var
Stop joining libcontainer executor process into the newly created task
container cgroup, to ensure that the cgroups are fully destroyed on
shutdown, and to make it consistent with other plugin processes.
Previously, executor process is added to the container cgroup so the
executor process resources get aggregated along with user processes in
our metric aggregation.
However, adding executor process to container cgroup adds some
complications with much benefits:
First, it complicates cleanup. We must ensure that the executor is
removed from container cgroup on shutdown. Though, we had a bug where
we missed removing it from the systemd cgroup. Because executor uses
`containerState.CgroupPaths` on launch, which includes systemd, but
`cgroups.GetAllSubsystems` which doesn't.
Second, it may have advese side-effects. When a user process is cpu
bound or uses too much memory, executor should remain functioning
without risk of being killed (by OOM killer) or throttled.
Third, it is inconsistent with other drivers and plugins. Logmon and
DockerLogger processes aren't in the task cgroups. Neither are
containerd processes, though it is equivalent to executor in
responsibility.
Fourth, in my experience when executor process moves cgroup while it's
running, the cgroup aggregation is odd. The cgroup
`memory.usage_in_bytes` doesn't seem to capture the full memory usage of
the executor process and becomes a red-harring when investigating memory
issues.
For all the reasons above, I opted to have executor remain in nomad
agent cgroup and we can revisit this when we have a better story for
plugin process cgroup management.
Copy the updated version of freeport (sdk/freeport), and tweak it for use
in Nomad tests. This means staying below port 10000 to avoid conflicts with
the lib/freeport that is still transitively used by the old version of
consul that we vendor. Also provide implementations to find ephemeral ports
of macOS and Windows environments.
Ports acquired through freeport are supposed to be returned to freeport,
which this change now also introduces. Many tests are modified to include
calls to a cleanup function for Server objects.
This should help quite a bit with some flakey tests, but not all of them.
Our port problems will not go away completely until we upgrade our vendor
version of consul. With Go modules, we'll probably do a 'replace' to swap
out other copies of freeport with the one now in 'nomad/helper/freeport'.
Operators commonly have docker logs aggregated using various tools and
don't need nomad to manage their docker logs. Worse, Nomad uses a
somewhat heavy docker api call to collect them and it seems to cause
problems when a client runs hundreds of log collections.
Here we add a knob to disable log aggregation completely for nomad.
When log collection is disabled, we avoid running logmon and
docker_logger for the docker tasks in this implementation.
The downside here is once disabled, `nomad logs ...` commands and API
no longer return logs and operators must corrolate alloc-ids with their
aggregated log info.
This is meant as a stop gap measure. Ideally, we'd follow up with at
least two changes:
First, we should optimize behavior when we can such that operators don't
need to disable docker log collection. Potentially by reverting to
using pre-0.9 syslog aggregation in linux environments, though with
different trade-offs.
Second, when/if logs are disabled, nomad logs endpoints should lookup
docker logs api on demand. This ensures that the cost of log collection
is paid sparingly.
Looks like the RecoverTask doesn't set taskHandle.logger field causing
a panic when the handle attempts to log (e.g. when Shutdown or Signaling
fails).
When a job has a task group network, this log line ends up being
misleading if you're trying to debug networking issues. We really only
care about this when there's no port map set, in which case we get the
error returned anyways.
driver.SetConfig is not appropriate for starting up reconciler
goroutine. Some ephemeral driver instances are created for validating
config and we ought not to side-effecting goroutines for those.
We currently lack a lifecycle hook to inject these, so I picked the
`Fingerprinter` function for now, and reconciler should only run after
fingerprinter started.
Use `sync.Once` to ensure that we only start reconciler loop once.
When running at scale, it's possible that Docker Engine starts
containers successfully but gets wedged in a way where API call fails.
The Docker Engine may remain unavailable for arbitrary long time.
Here, we introduce a periodic reconcilation process that ensures that any
container started by nomad is tracked, and killed if is running
unexpectedly.
Basically, the periodic job inspects any container that isn't tracked in
its handlers. A creation grace period is used to prevent killing newly
created containers that aren't registered yet.
Also, we aim to avoid killing unrelated containters started by host or
through raw_exec drivers. The logic is to pattern against containers
environment variables and mounts to infer if they are an alloc docker
container.
Lastly, the periodic job can be disabled to avoid any interference if
need be.
This commit introduces support for configuring mount propagation when
mounting volumes with the `volume_mount` stanza on Linux targets.
Similar to Kubernetes, we expose 3 options for configuring mount
propagation:
- private, which is equivalent to `rprivate` on Linux, which does not allow the
container to see any new nested mounts after the chroot was created.
- host-to-task, which is equivalent to `rslave` on Linux, which allows new mounts
that have been created _outside of the container_ to be visible
inside the container after the chroot is created.
- bidirectional, which is equivalent to `rshared` on Linux, which allows both
the container to see new mounts created on the host, but
importantly _allows the container to create mounts that are
visible in other containers an don the host_
private and host-to-task are safe, but bidirectional mounts can be
dangerous, as if the code inside a container creates a mount, and does
not clean it up before tearing down the container, it can cause bad
things to happen inside the kernel.
To add a layer of safety here, we require that the user has ReadWrite
permissions on the volume before allowing bidirectional mounts, as a
defense in depth / validation case, although creating mounts should also require
a priviliged execution environment inside the container.
Without passing the network isolation configuration to the executor,
java tasks are not placed in the same network namespace as the other
processes in their task group, which breaks Consul Connect.
The docker creation API calls may fail with http errors (e.g. timeout)
even if container was successfully created.
Here, we force remove container if we got unexpected failure. We
already do this in some error handlers, and this commit updates all
paths.
I stopped short from a more aggressive refactoring, as the code is ripe
for refactoring and would rather do that in another PR.
This handles a bug where we may start a container successfully, yet we
fail due to retries and startContainer not being idempotent call.
Here, we ensure that when starting a container fails with 500 error,
the retry succeeds if container was started successfully.
hclspec.NewLiteral does not quote its values, which caused `3m` to be
parsed as a nonsensical literal which broke the plugin loader during
initialization. By quoting the value here, it starts correctly.
Seth Hoenig