Open manishtomar opened 9 years ago
Right now in worker, otter has 2 ids associated with each request:
otter-request-id
header when talking to Nova/CLB.otter-transaction-id
header when talking to Nova/CLB. Scheduled policy works differently. Each schedule run has an ID that can be sent as transaction id header to HTTP request sent during scheduled policy executions in that run. Please note that there can be different tenant's different groups getting executed in same schedule run. This ID would be different in convergence. Each converge run can have ID that can correspond to this ID. Does above changes sound reasonable?
I have no idea why I missed the second comment. Yes that sounds reasonable.
Update: we might want to prepend an "X-" to all those header names for non-standard names? I think that was deprecated (https://en.wikipedia.org/wiki/List_of_HTTP_header_fields#cite_note-2) but the reason was that it made things difficult when non-standard became standard, but if 'otter' is in the name that is probably not a danger.
Update: the per-request ID is done. The transaction ID has not been done yet, since at least with the converger, it may not be part of a transaction, but a convergence run.
Add request id header from otter's transaction id when sending request to Nova/CLB. This allows us to correlate requests from otter to Nova.