search

radical-collaboration / QCArchive

2 stars 0 forks source link

Heart beat management from RP to QCA #12

Open mturilli opened 5 years ago

mturilli commented 5 years ago

To be a requirement.

mturilli commented 5 years ago

REST interface on RP-side is not required anymore

dgasmith commented 5 years ago

We will still need heartbeats from RP to double check that it is still alive. This is a work in progress in QCA, please give us a few weeks to get this properly implemented.

andre-merzky commented 5 years ago

I assume that you'll define a REST method you want us to call on regular intervals? Well, we'll be pulling for tasks at regular intervals, that could naively be used as a heartbeat, too. But I also don't mind doing that cleanly and calling a separate REST method, of course.

dgasmith commented 5 years ago

Yes to a heartbeat REST method.

We could set it up as a task pull counts as a heartbeat as well. Do you have thoughts/experience convolving heartbeats/task pulls together? Technically easy to do, but having a clean separation does have upsides.

andre-merzky commented 5 years ago

Yeah, I don't mind clean separation - we do the same in RP. Not for semantic reasons though, but for performance and reliability: I don't want to have heartbeats stuck in large task queues... But that's only the case because we intermediate via a DB right now, and cannot evaluate the actual connection as heartbeat...

mturilli commented 5 years ago

@dgasmith to implement the dedicated interface in at least 1 month. Meanwhile QCA will do manual boo-keeping of suspended jobs follow a RP failure.

mturilli commented 5 years ago

Ongoing