Open mwes opened 3 years ago
IP has been updated to use new TACC endpoint to address #252. This issue, however, will require more changes based on the Slack conversation that went on between @mwes and @mwvaughn on 10/21/2020. As mentioned in the conversation, #252 is in a good state for @mwes to use for milestone 2.10. @mwes will continue to help other users debug the state of an experiment execution until #257 is resolved.
New workflow that this issue will need to build off of:
request_id
that matches to a experiment_reference_url_for_xplan
. request_id
-> execution_id
execution_id
can then get passed into TACCGoAccessor.get_status_of_experiment to get the status of an experiment execution. Information returned from this function, should be reported back to the user in a human readable form.
The current request experiment execution path should return a JSON response that follows this template:
What this is saying is that the request was received and accepted by the reactor, and assigned an execution id
7LYepwB1ewLb3
which you can see under theresult
field.This does not provide visibility into any possible errors in that execution. To do that, we need to check the execution status. A GET request to the following:
https://api.sd2e.org/actors/v2/control-annotator.prod/executions/7LYepwB1ewLb3?x-nonce=$NONCE
will retrieve that. The nonce has been provided to you in a side-channel. This returns JSON as well:
Note the
exit_code
andstatus
fields:after being submitted for execution, the reactor will process and transition to the completed state when done.
When
status
is "COMPLETE" theexit_code
will be valid.For non-zero exit codes, we can pull (and show) logs for the execution via:
https://api.sd2e.org/actors/v2/control-annotator.prod/executions/7LYepwB1ewLb3/logs?x-nonce=$NONCE
This will allow visibility/clarity into reactor executions that succeed or fail, and if they fail, what the nature of the error was.