conjurelabs/conjure-worker

queue & directly-called task workers

conjure-worker's Issues

create lambda to check if instances are still running

verify updating containers does not overwrite row

in order to bill people, updates to containers may need to be separate rows.

or, could make a separate table to track multiple run spans

turn on jobs on lambda

missing `task` when stopping non-existent task

0|conjure- | github.container.stop -->  stopping github container (ConjureLabs/mock-web-repo --> prod-2)
0|conjure- | container stop -->  stopping container
0|conjure- | container stop -->  stopping task
0|conjure- | github.container.stop -->  { MissingRequiredParameter: Missing required key 'task' in params
0|conjure- |     at ParamValidator.fail (/home/ubuntu/conjure-worker/node_modules/conjure-core/node_modules/aws-sdk/lib/param_validator.js:50:37)
0|conjure- |     at ParamValidator.validateStructure (/home/ubuntu/conjure-worker/node_modules/conjure-core/node_modules/aws-sdk/lib/param_validator.js:61:14)
0|conjure- |     at ParamValidator.validateMember (/home/ubuntu/conjure-worker/node_modules/conjure-core/node_modules/aws-sdk/lib/param_validator.js:88:21)
0|conjure- |     at ParamValidator.validate (/home/ubuntu/conjure-worker/node_modules/conjure-core/node_modules/aws-sdk/lib/param_validator.js:34:10)
0|conjure- |     at Request.VALIDATE_PARAMETERS (/home/ubuntu/conjure-worker/node_modules/conjure-core/node_modules/aws-sdk/lib/event_listeners.js:125:42)
0|conjure- |     at Request.callListeners (/home/ubuntu/conjure-worker/node_modules/conjure-core/node_modules/aws-sdk/lib/sequential_executor.js:105:20)
0|conjure- |     at callNextListener (/home/ubuntu/conjure-worker/node_modules/conjure-core/node_modules/aws-sdk/lib/sequential_executor.js:95:12)
0|conjure- |     at /home/ubuntu/conjure-worker/node_modules/conjure-core/node_modules/aws-sdk/lib/event_listeners.js:85:9
0|conjure- |     at finish (/home/ubuntu/conjure-worker/node_modules/conjure-core/node_modules/aws-sdk/lib/config.js:320:7)
0|conjure- |     at /home/ubuntu/conjure-worker/node_modules/conjure-core/node_modules/aws-sdk/lib/config.js:338:9
0|conjure- |   message: 'Missing required key \'task\' in params',
0|conjure- |   code: 'MissingRequiredParameter',
0|conjure- |   time: 2018-05-28T23:30:53.934Z }

add heartbeat to update container flow

Fix queues so they don't hang

should not fully rebuild if PR is closed and then re-opened

Secure private instances

A private instance is public if someone knows the host and port. Though they are ephemeral, thus private branches leaking is less likely, eventually someone would figure out the pattern and brute-force search the servers until they find all exposed ports. So, we need to wrap private instances (or all?) with a layer that rejects any request w/o the right headers.Or do not use ports at all? And use some sort of internal routing system?

lambda to check for pending container that is orphaned

if a container goes into pending state, kicks off a worker, and the worker throws before changing the state to spinning up, then that container will never get caught by the lambda check for stale heartbeats.

could possibly check against redis for what's in the queue, and see what either needs to be requeued or set as failed.

merging a pr did not wipe out gh comment

handle pending billing where billing plan is not longer active

If a user removes card, or billing plan is for whatever reason deactivated, before billing job happens, then the user's org should be frozen until billing is finished, and emails should go out to account on record.

bill users on some sort of cron

manual for now?

should wipe ecr record after a PR closes

Deal with fast webhook events

what happens if a user updates update updates, and all the events get queued?

Possibly certain events should cancel?