Igor Ponomarev (3b129806) at 18 Mar 16:04

I added hostname filtering in the device type overview table to my MR. Can you give it a test? (there should be a Search button on top of the table)

Igor Ponomarev (ec6c8b12) at 18 Mar 13:58

... and 6 more commits

Implicit device tag sounds like an over complication. Why not simply allow submitting job for a specific device.

I am reworking table searching in !2546. It will have a nice GUI to filter the table results.

I am reworking the tables search in !2546. It will have a nice GUI to filter the table results.

Rémi Duraffort (64f5f167) at 18 Mar 07:48

Rémi Duraffort (db3394e8) at 18 Mar 07:48

... and 1 more commit

Per the comments and the doc, the lava-role list should print the roles rather than the current job IDs.

The problem is that the lava-group script prints \t<job_id>\t<role> with the additional \t prefix. This doesn't match the example output in the doc, and I don't see how the prefix is used. It is more likely copied from the previous log line.

lava-role uses the lava-group output for listing the roles.

lava-group | cut -f2

Internally, lava-role is the only user of the lava-group output. Removing the prefix fixes the lava-role list API and shouldn't introduce any internal regressions.

Although I made the equivalent changes in the lava_multi_node_make_aliased_hosts API as well, the API is not used anywhere.

The change should be noted in the release note since it may impact users that rely on the current lava-group output for getting job ID and role. lava-self and lava-role should be used for these cases though.

Signed-off-by: Chase Qi [email protected]

Per the comments and the doc, the lava-role list should print the roles rather than the current job IDs.

The problem is that the lava-group script prints \t<job_id>\t<role> with the additional \t prefix. This doesn't match the example output in the doc, and I don't see how the prefix is used. It is more likely copied from the previous log line.

lava-role uses the lava-group output for listing the roles.

lava-group | cut -f2

Internally, lava-role is the only user of the lava-group output. Removing the prefix fixes the lava-role list API and shouldn't introduce any internal regressions.

Although I made the equivalent changes in the lava_multi_node_make_aliased_hosts API as well, the API is not used anywhere.

The change should be noted in the release note since it may impact users that rely on the current lava-group output for getting job ID and role. lava-self and lava-role should be used for these cases though.

Signed-off-by: Chase Qi [email protected]

Chase Qi (64f5f167) at 18 Mar 07:33

Chase Qi (5df529bf) at 17 Mar 06:42

... and 1 more commit

Action item for !3176 (comment 3133820023)

Typically, when we submit a job, LAVA randomly selects a device that meets the tag requirements. However, when we later want to rerun the job on the exact same board, this isn't possible unless:

• The tag is unique to that device
• Other boards are busy
• We manually add a device name tag to that device's configuration (which is tedious for template configurations, and furthermore, these boards are sometimes not owned by me)

I'd like LAVA to have an implicit scheduling mechanism: even if a device is not configured with a device name tag, when a user specifies a device name in the job definition, the job should be directly assigned to that device. This would be especially useful for reruns when we need to verify issues on the same board.

Ben Copeland (c2de6c52) at 16 Mar 08:45

I guess the commit msg is a bit confusing. The medium runner tag is available to lava namespace but not personal namespace.The small tag available for both.

Rémi Duraffort (9c62ff72) at 16 Mar 08:36

... and 1 more commit

Rémi Duraffort (24313c0a) at 16 Mar 08:36

saas-linux-medium-arm64 is Premium and Ultimate only. LAVA can use it probably because it is under the OSS policy.

Switch to saas-linux-small-arm64 so the personal forks can run the test and build jobs. Per the spec below, it should be good enough for our jobs.

https://docs.gitlab.com/ci/runners/hosted_runners/linux/#machine-types-available-for-linux---arm64

Signed-off-by: Chase Qi [email protected]

saas-linux-medium-arm64 is Premium and Ultimate only. LAVA can use it probably because it is under the OSS policy.

Switch to saas-linux-small-arm64 so the personal forks can run the test and build jobs. Per the spec below, it should be good enough for our jobs.

https://docs.gitlab.com/ci/runners/hosted_runners/linux/#machine-types-available-for-linux---arm64

Signed-off-by: Chase Qi [email protected]