Also add some debug log lines for this test, because it doesn't make sense
for the allocation to be complete yet a task in the allocation to be not
started yet, which is what the test failures are implying.
Originally this test relied on Job 1 blocking Job 2 until Job 1 had a
terminal *ClientStatus.* Job 2 ensured it would get blocked using 2
mechanisms:
1. A constraint requiring it is placed on the same node as Job 1.
2. Job 2 would require all unreserved CPU on the node to ensure it would
be blocked until Job 1's resources were free.
That 2nd assertion breaks if *any previous job is still running on the
target node!* That seems very likely to happen in the flaky world of our
e2e tests. In fact there may be some jobs we intentionally want running
throughout; in hindsight it was never safe to assume my test would be
the only thing scheduled when it ran.
*Ports to the rescue!* Reserving a static port means that both Job 2
will now block on Job 1 being terminal. It will only conflict with other
tests if those tests use that port *on every node.* I ensured no
existing tests were using the port I chose.
Other changes:
- Gave job a bit more breathing room resource-wise.
- Tightened timings a bit since previous failure ran into the `go test`
time limit.
- Cleaned up the DumpEvals output. It's quite nice and handy now!
Keeps failing in the nightly e2e test with unhelpful output like:
```
Failed
=== RUN TestOverlap
overlap_test.go:92: Followup job overlap93ee1d2b blocked. Sleeping for the rest of overlap48c26c39's shutdown_delay (9.2/10s)
overlap_test.go:105: 1500/2000 retries reached for github.com/hashicorp/nomad/e2e/overlap.TestOverlap (err=timed out before an allocation was found for overlap93ee1d2b)
overlap_test.go:105: timeout: timed out before an allocation was found for overlap93ee1d2b
--- FAIL: TestOverlap (38.96s)
```
I have not been able to replicate it in my own e2e cluster, so I added
the EvalDump helper to add detailed eval information like:
```
=== RUN TestOverlap
1/1 Job overlap7b0e90ec Eval c38c9919-a4f0-5baf-45f7-0702383c682a
Type: service
TriggeredBy: job-register
Deployment:
Status: pending ()
NextEval:
PrevEval:
BlockedEval:
-- No placement failures --
QueuedAllocs:
SnapshotIdx: 0
CreateIndex: 96
ModifyIndex: 96
...
```
Hopefully helpful when debugging other tests as well!
* test: simplify overlap job placement logic
Trying to fix#14806
Both the previous approach as well as this one worked on e2e clusters I
spun up.
* simplify code flow
* test: add e2e for non-overlapping placements
Followup to #10446
Fails (as expected) against 1.3.x at the wait for blocked eval (because
the allocs are allowed to overlap).
Passes against 1.4.0-beta.1 (as expected).
* Update e2e/overlap/overlap_test.go
Co-authored-by: James Rasell <jrasell@users.noreply.github.com>
* cleanup: refactor MapStringStringSliceValueSet to be cleaner
* cleanup: replace SliceStringToSet with actual set
* cleanup: replace SliceStringSubset with real set
* cleanup: replace SliceStringContains with slices.Contains
* cleanup: remove unused function SliceStringHasPrefix
* cleanup: fixup StringHasPrefixInSlice doc string
* cleanup: refactor SliceSetDisjoint to use real set
* cleanup: replace CompareSliceSetString with SliceSetEq
* cleanup: replace CompareMapStringString with maps.Equal
* cleanup: replace CopyMapStringString with CopyMap
* cleanup: replace CopyMapStringInterface with CopyMap
* cleanup: fixup more CopyMapStringString and CopyMapStringInt
* cleanup: replace CopySliceString with slices.Clone
* cleanup: remove unused CopySliceInt
* cleanup: refactor CopyMapStringSliceString to be generic as CopyMapOfSlice
* cleanup: replace CopyMap with maps.Clone
* cleanup: run go mod tidy
In the event a single test fails to clear up properly after
itself, all other tests will fail as they attempt to create ACL
policies with the same names. This change ensures they use unique
ACL names, so when a single test fails, it is easy to identify it
is a problem with the test rather than the suite.
The rewrite refactors the suite to use the new style along with
other recent testing improvements. In order to ensure the spread
tests do not impact each other, there is new cleanup functionality
to ensure both the job and allocations are removed from state
before the test exits completely.
The NSD checks tests were racey, whereby the check may not have
been triggered by the time it was queried. This change wraps the
check so it can account for this.
This removes the current ACL expiration GC section in order to get
the tests passing and allow more time to investigate the test. I
have full confidence the feature is working as expected and have
tested extensively locally.
In order to add an E2E test to cover token expiration, the server
config has been updated to include a low minimum allowed TTL
value. For ease of reading, the max value is also set.
This adds a new ACL test suite to the e2e framework which includes
an initial test for ACL roles. The ACL test includes a helper to
track and clean created Nomad resources which keeps the test
cluster clean no matter if the test fails early or not.
* core: allow pause/un-pause of eval broker on region leader.
* agent: add ability to pause eval broker via scheduler config.
* cli: add operator scheduler commands to interact with config.
* api: add ability to pause eval broker via scheduler config
* e2e: add operator scheduler test for eval broker pause.
* docs: include new opertor scheduler CLI and pause eval API info.
The nightly playwright tests are currently failing because of a
mismatch between the expected version of Chromium and what's in the
container image. Unfortunately the previous specific tag we were using
for the container image is no longer tagged on the registry. With some
testing, I was able to find an image tag that results in a good run.
This test exercises upgrades between 0.8 and Nomad versions greater
than 0.9. We have not supported 0.8.x in a very long time and in any
case the test has been marked to skip because the downloader doesn't
work.
We moved off the old provisioning process for nightly E2E to one driven
entirely by Terraform quite a while back now. We're in the slow
process of removing the framework code for this test-by-test, but this
chunk of code no longer has any callers.
We don't need the absolute path for any of the commands in this script
so long as we `cd` into the source directory path. Doing this removes
the need for weird platform-specific tricks we have to do with
realpath vs GNU realpath.
The E2E test runner is running from the root of the Nomad
repository. Make this run independent of the working directory for
convenience of developers and the test runner.
The CSI plugin allocations take a while to be marked healthy,
sometimes causing E2E test flakes during the setup phase of the
tests. There's nothing CSI specific about marking plugin allocs
healthy, as the plugin supervisor hook does all the fingerprinting in
the postrun hook (the prestart hook just makes a couple of empty
directories). The timeouts we're seeing may be because of where we're
pulling the images from; most our jobs pull from a CDN-backed public
registry whereas these are pulling from ECR. Set a 1min timeout for
these to make sure we have enough time to pull the image and start the
task.
Scripts for running playwright tests in a Docker container that has
chromium and webkit preinstalled. Includes a basic smoke test for
authentication so that we can be sure the test rig is working
end-to-end. Wiring this up in CI will be in an upcoming PR.
Our E2E test environment is deployed with mTLS, but it's impractical
for us to use mTLS in headless browsers for automated testing (or even
in manual testing). Provide certificates for proxying the web UI via
Nginx. This proxy uses client certs for proxying to the HTTP endpoint
and a self-signed cert for the browser-facing endpoint. We can accept
certificate errors in the automated tests we'll be adding in the next
step of this work.
While working on infrastructure for testing the UI in E2E, we needed
to upgrade the certificate provider. Performing a provider upgrade via
the TF `init -upgrade` brought in updates for the file and AWS
providers as well. These updates include deprecating the use of
`sensitive_content` fields, removing CA algorithm parameters that can
be inferred from keys, and removing the requirement to manually
specify AWS assume role parameters in the provider config if they're
available in the calling environment's AWS config file (as they are
via doormat or our E2E environment).
This test has a failure that's happening only occassionally and not
very reproducibly. Print out the allocation status on test failure so
that we can do some post-mortum debugging of the test on nightly.
Many of our scripts have a non-portable interpreter line for bash and
use bash-specific variables like `BASH_SOURCE`. Update the interpreter
line to be portable between various Linuxes and macOS without
complaint from posix shell users.
This changeset fixes two sources of flakiness in the event stream test.
First, the stream request gets the event *closest* to the index, not
the exact match. Although events are written before raft entries
they're written asynchronously, so it's possible to race and get a
raft index from this query higher than the current head of the event
buffer. Ensure the job is running before we try to get the index, so
that we've given the event enough time to land in the buffer.
Second, the assertion that the found index is greater than the start
index is only true if the `PlanResult` event manages to land before we
do the second registration. Although it should now with the first fix
above, it's not a correct assertion for what we're testing.
The oversubscription test expects an output that requires the client
has polled the task for stats at least once. Wait long enough to
ensure that we've polled the stats before failing the test.
Some tests may chose to deregister jobs to check Nomad cleanup
logic, however, it is still possible for the test to fail and exit
before this is hit. This therefore adds a cancellable cleanup func
which can be deferred, using context to control whether it gets
run or not.
This change modifies the template task runner to utilise the
new consul-template which includes Nomad service lookup template
funcs.
In order to provide security and auth to consul-template, we use
a custom HTTP dialer which is passed to consul-template when
setting up the runner. This method follows Vault implementation.
Co-authored-by: Michael Schurter <mschurter@hashicorp.com>
Tear down the volume-consuming job between subtests, rather than after
all the tests are complete. For good measure, use a different ID for
the volume-consuming job as well.
* Wait longer for node to go down in disconnected clients test.
The existing helper only waits 10s, but there's a jitter on heartbeats
that we need to account for. Wait for 30s for node to go down to give
us plenty of room
* Port disconnected clients to stdlib-style test
Concurrent E2E runs can collide when provisioning policies on HCP
Consul and HCP Vault. Namespace these by the test run name, as we do
for most everything else.
Our E2E "framework" has a bunch of features around test discovery and
standing up infra that were never completed or fully used, and we
ended up building out a large test suite that ignored all that in lieu
of Terraform-provided infrastructure for the last couple years.
This changeset is a proposal (and demonstration) for gradually
migrating our E2E tests off the framework code so that developers can
write fairly ordinary golang stdlib testing tests.
This test exercises the behavior of clients that become disconnected
and have their allocations replaced. Future test cases will exercise
the `max_client_disconnect` field on the job spec.
* Use unix:// prefix for CSI_ENDPOINT variable by default
* Some plugins have strict validation over the format of the
`CSI_ENDPOINT` variable, and unfortunately not all plugins
agree. Allow the user to override the `CSI_ENDPOINT` to workaround
those cases.
* Update all demos and tests with CSI_ENDPOINT
The `ConnectACLsE2ETest` checks that the SI tokens have been properly
cleaned up between tests, but following the change to use HCP the
previous `Connect` test suite will often have SI tokens that haven't
been cleaned up by the time this test suite runs. Wait for the SI
tokens to be cleaned up at the start of the test to ensure we have a
clean state.
Use HCP Consul and HCP Vault for the Consul and Vault clusters used in E2E testing. This has the following benefits:
* Without the need to support mTLS bootstrapping for Consul and Vault, we can simplify the mTLS configuration by leaning on Terraform instead of janky bash shell scripting.
* Vault bootstrapping is no longer required, so we can eliminate even more janky shell scripting
* Our E2E exercises HCP, which is important to us as an organization
* With the reduction in configurability, we can simplify the Terraform configuration and drop the complicated `provision.sh`/`provision.ps1` scripts we were using previously. We can template Nomad configuration files and upload them with the `file` provisioner.
* Packer builds for Linux and Windows become much simpler.
tl;dr way less janky shell scripting!
This is a followup to having tests run in serial in CI.
The e2e package isn't in CI, but lets use the helper anyway
so we can setup semgrep rules covering the entire repository.
The RPC for listing volume snapshots requires a plugin ID. Update the
`volume snapshot list` command to find the specific plugin from the
provided prefix.
If any E2E test hangs, it'll eventually timeout and panic, causing the
all the remaining tests to fail. External commands should use a short
context whenever possible so we can fail the test quickly and move on
to the next test.
The `TestRescheduleProgressDeadlineFail` E2E test failed during test
cleanup because the error message "progress deadline expired" that it
emits when we stop the job does not match the one expected from
monitoring the `job stop` command. Update the `StopJob` helper to
tolerate this use case as well.
The `Metrics` suite uses prometheus to scrape Nomad metrics so that
we're testing the full user experience of extracting metrics from
Nomad. With the addition of mTLS, we need to make sure prometheus also
has mTLS configuration because the metrics endpoint is protected.
Update the Nomad client configuration and prometheus job to bind-mount
the client's certs into the task so that the job can use these certs
to scrape the server. This is a temporary solution that gets the job
passing; we should give the job its own certificates (issued by
Vault?) when we've done some of the infrastructure rework we'd like.
The AWS EBS plugin appears to use the name field of the volume as an
idempotency token that persists across the entire AWS account, not
just the plugin lifespan.
Also fix the regex for the volume ID, which was originally taken from
the job ID regex but isn't actually the same. This hasn't failed tests
for us because we've always passed in the same volume ID.
With mTLS enabled, using `curl` in a bash script for validation
involves having to configure arguments to `curl` based on whether or
not the test infrastructure is using mTLS, whether ACLs are enabled,
etc. Use the new `operator api` command instead to pick up the client
configuration from the test environment automatically.
PR #11550 changed the job stop exit behaviour when monitoring the
deployment. When stopping a job, the deployment becomes cancelled
and therefore the CLI now exits with status code 1 as it see this
as an error.
This change adds a new utility e2e function that accounts for this
behaviour.
This change modifies the Nomad job register and deregister RPCs to
accept an updated option set which includes eval priority. This
param is optional and override the use of the job priority to set
the eval priority.
In order to ensure all evaluations as a result of the request use
the same eval priority, the priority is shared to the
allocReconciler and deploymentWatcher. This creates a new
distinction between eval priority and job priority.
The Nomad agent HTTP API has been modified to allow setting the
eval priority on job update and delete. To keep consistency with
the current v1 API, job update accepts this as a payload param;
job delete accepts this as a query param.
Any user supplied value is validated within the agent HTTP handler
removing the need to pass invalid requests to the server.
The register and deregister opts functions now all for setting
the eval priority on requests.
The change includes a small change to the DeregisterOpts function
which handles nil opts. This brings the function inline with the
RegisterOpts.
Add a new hostname string parameter to the network block which
allows operators to specify the hostname of the network namespace.
Changing this causes a destructive update to the allocation and it
is omitted if empty from API responses. This parameter also supports
interpolation.
In order to have a hostname passed as a configuration param when
creating an allocation network, the CreateNetwork func of the
DriverNetworkManager interface needs to be updated. In order to
minimize the disruption of future changes, rather than add another
string func arg, the function now accepts a request struct along with
the allocID param. The struct has the hostname as a field.
The in-tree implementations of DriverNetworkManager.CreateNetwork
have been modified to account for the function signature change.
In updating for the change, the enhancement of adding hostnames to
network namespaces has also been added to the Docker driver, whilst
the default Linux manager does not current implement it.
This allows us to spin up e2e clusters with mTLS configured for all HashiCorp services, i.e. Nomad, Consul, and Vault. Used it for testing #11089 .
mTLS is disabled by default. I have not updated Windows provisioning scripts yet - Windows also lacks ACL support from before. I intend to follow up for them in another round.
Target all e2e datacenters for system and sysbatch e2e tests. They
require that the system jobs run on all linux clients.
However, the jobs currenly only target `dc1` datacenter, but the nightly
e2e cluster has 4 clients spread in `dc1` and `dc2` datacenters, causing
the tests to fail.
I missed this problem in e2e dev cluster because it only used a single
dc1 datacenter.
This PR implements a new "System Batch" scheduler type. Jobs can
make use of this new scheduler by setting their type to 'sysbatch'.
Like the name implies, sysbatch can be thought of as a hybrid between
system and batch jobs - it is for running short lived jobs intended to
run on every compatible node in the cluster.
As with batch jobs, sysbatch jobs can also be periodic and/or parameterized
dispatch jobs. A sysbatch job is considered complete when it has been run
on all compatible nodes until reaching a terminal state (success or failed
on retries).
Feasibility and preemption are governed the same as with system jobs. In
this PR, the update stanza is not yet supported. The update stanza is sill
limited in functionality for the underlying system scheduler, and is
not useful yet for sysbatch jobs. Further work in #4740 will improve
support for the update stanza and deployments.
Closes#2527
As we moved to using `-detach` for registering jobs, we should wait
until allocs and deployments are created before asserting their
properties.
Fixing `TestNodeDrainIgnoreSystem` and `TestRescheduleProgressDeadlineFail` tests as they seem particularly flaky, failing 9 and 7 times (respectively) in the last two weeks.
Pick up 15d39f0dee but for RegisterFromJobspec:
> This PR changes the e2e helper thingy to set -detach option
> when registering a job with the CLI instead of the API. This is
> necessary for jobs which never become healthy, as the deployment
> never finishes for failing jobs and the command never returns,
> causing the test to timeout after 10 minutes.
This case occurs in TestVaultSecrets
This PR changes the e2e helper thingy to set -detach option
when registering a job with the CLI instead of the API. This is
necessary for jobs which never become healthy, as the deployment
never finishes for failing jobs and the command never returns,
causing the test to timeout after 10 minutes.
Seth Hoenig