Wrong expected output when a timeout occurs

[LucaRUnipi]

I found a problem in yRCA output when the invoking service emits as log the template corresponding to the client_timeout proposed in /data/templates/chaos-echo.yml

The problem has been found testing Online Boutique application, where I intentionally put a sleep time in the responding service (adservice). yRCA's output was the following: "Found no failure cascade to the considered event", using the client_timeout template as error to explain.

• frontend ---> adservice (waits 3 sec)
• frontend logs "Failing to contact adservice (request_id: ...). Root cause: (Deadline exceeded)"

I think that yRCA should recognize that frontend had an internal error due to the timeout happened in adservice.

[jsoldani]

As for timeouts expiring, the Prolog explainer of yRCA currently handles two situations:

• explicitly logged failures of invoked services, due to which the timeouts in invokers expired (lines 14-26) or
• requests sent but never received/handled by invoked services (lines 77-87)

The described situation seem to relate to the first case (lines 14-26), for which yRCA should consider the base case of a service taking "too long" to answer to its invokers (same case, but no failure in the invoked service while interaction is going on), e.g.,

causedBy(log(SI,I,T,E,M,Sev),[X],SI) :-                        
    (E=errorFrom(SJ,Id);E=timeout(SJ,Id)),
    failedInteraction(Id,(SI,I),(SJ,J),Ts,Te),
    \+ (log(SJ,J,U,_,_,SevJ), lte(SevJ,warning), Ts =< U, U =< Te),
    X=log(SI,I,T,E,M,Sev).

(which would cover also the error responses due to processing "bad requests" of invokers)