LAVA multi-node support allows users to use LAVA to schedule, synchronize and combine the results from tests that span multiple targets. Jobs can be arranged as groups of devices (of any type) and devices within a group can operate independently or use the MultiNode API to communicate with other devices in the same group during tests.
Within a MultiNode group, devices of the same device type are assigned a role
count of devices to include into that role. Role labels must be
unique across the entire MultiNode job. Each role has a
device_type and any
number of roles can have the same
device_type. Each role can be assigned
Once roles are defined, actions (including test images and test definitions) can be marked as applying to specific roles (if no role is specified, all roles use the action).
If insufficient boards exist to meet the combined requirements of all the roles specified in the job, the job will be rejected.
If there are not enough idle boards of the relevant types to meet the combined requirements of all the roles specified in the job, the job waits in the Submitted queue until all devices can be allocated.
Each test job is put into a MultiNode group and basic information about other jobs in that group will be available in each test shell using the MultiNode API.
Using LAVA MultiNode synchronization¶
Once each board has booted the test image, the MultiNode API will also be available for use within each test definition using scripts placed into the default PATH by the LAVA overlay.
Unless two or more roles use the MultiNode API to synchronize operations at some point within the test job submission, the test jobs will start at the same time but run independently. Even if the test jobs in a MultiNode group are identical, the time taken to download, deploy and boot into the test shell will vary. There is no guarantee that a service will be available for another role in the MultiNode group unless the test writer uses the synchronization primitives in the MultiNode API. This also applies to tests where one role needs to send data (like an IP address) to another role. One of the first tasks for many MultiNode test jobs is to synchronize specific roles.
In all roles in this MultiNode group:
- lava-sync server
To synchronize only specific roles, send a specific string using lava-send and make the other role use lava-wait with that same string. Then send another message from the waiting role and make the first role wait for the second message.
In the role acting as a server:
- lava-send server - lava-wait client
In the role acting as a client:
- lava-wait server - lava-send client
If one role is essential to all other roles in the test job, for example if a role has to install and configure a server which is to be contacted by other roles within the group, mark that role as essential. When the job(s) marked with the essential role fail, all test jobs in the MultiNode group will terminate.
To make your test job submissions more portable, it is recommended to use inline test definitions when calling the MultiNode API from the test shell. All MultiNode API calls can also be executed from custom scripts although this can make things harder to debug.
MultiNode synchronization calls will exit non-zero if the attempt times out or fails in some other way. The test shell definition will then exit at this point.
It is not recommended to wrap MultiNode synchronization calls in calls to
lava-test-case because if the API call fails,
report a fail result but the test definition itself will continue as if the
synchronization succeeded. The synchronization calls themselves will create
results based on the operation requested.
Each call to lava-send, lava-sync, lava-wait or
lava-wait-all will generate a test case with a
prefix in the current test suite of the results for this test job. If
the synchronization completes within the timeout, the result will be a
pass. If the attempt to synchronize times out, the result will be a
- lava-wait server - lava-send client
Would generate test case results like
LAVA MultiNode timeout behavior¶
The submitted YAML includes a timeout value - in single node LAVA, this is applied to each individual action executed on the device under test (not for the entire job as a whole). i.e. the default timeout can be smaller than any one individual timeout used in the YAML or internally within LAVA.
In MultiNode LAVA, this timeout is also applied to individual polling operations, so an individual lava-sync or a lava-wait will fail on any node which waits longer than the default timeout. The node will receive a failure response.
Marking some roles as essential - if your test job involves a long running server and clients, marking the server as essential allows the client test jobs to fail early instead of waiting for a long timeout.
Recommendations on timeouts for MultiNode¶
MultiNode operations have implications for the timeout values used in YAML submissions. If one of the synchronization primitives times out, the sync will fail and the job itself will then time out. One reason for a MultiNode job to timeout is if one or more boards in the group failed to boot the test image correctly. In this situation, all the other boards will continue until the first synchronization call is made in the test definition for that board.
The time limit applied to a synchronization primitive starts when the board makes the first request to the Coordinator for that sync. Slower boards may well only get to that point in the test definition after faster devices (especially KVM devices) have started their part of the sync and timed out themselves.
Always review the protocol timeout and job timeouts in the YAML submission. Excessive timeouts would prevent other jobs from using boards where the waiting jobs have already failed due to a problem elsewhere in the group. If timeouts are too short, jobs will fail unnecessarily.
Running a server on the device-under-test¶
If this server process runs as a daemon, the test definition will need to define something for the device under test to actually do or it will simply get to the end of the tests and reboot. For example, if the number of operations is known, would be to batch up commands to the daemon, each batch being a test case. If the server program can run without being daemonised, it would need to be possible to close it down at the end of the test (normally this is the role of the sysadmin in charge of the server box itself).
Making use of third party servers¶
A common part of a MultiNode setup is to download components from third party servers but once the test starts, latency and connectivity issues could interfere with the tests.
Using wrapper scripts¶
Wrapper scripts make it easier to test your definitions before submitting to LAVA. The wrapper lives in a VCS repository which is specified as one of the testdef_repos and will be available in the same directory structure as the original repository. A wrapper script also helps the tests to fail early instead of trying to do the rest of the tests.