|Summary:||if getReturnValue().setSucceeded(true); is called before exception is thrown command is presented as successful.|
|Product:||Red Hat Enterprise Virtualization Manager||Reporter:||Martin Mucha <mmucha>|
|Component:||ovirt-engine||Assignee:||Ravi Nori <rnori>|
|Status:||CLOSED CURRENTRELEASE||QA Contact:||Martin Mucha <mmucha>|
|Version:||3.5.0||CC:||gklein, iheim, lpeer, michal.skrivanek, oourfali, pstehlik, rbalakri, Rhev-m-bugs, rnori, sherold, yeylon, yzaslavs|
|Fixed In Version:||ovirt-engine-3.5.0_beta||Doc Type:||Bug Fix|
|Doc Text:||Story Points:||---|
|Last Closed:||2015-02-17 17:07:32 UTC||Type:||Bug|
|oVirt Team:||Infra||RHEL 7.3 requirements from Atomic Host:|
|Cloudforms Team:||---||Target Upstream Version:|
|Bug Depends On:|
|Bug Blocks:||1142923, 1156165|
Description Martin Mucha 2014-06-09 12:15:16 UTC
Description of problem: if any Command calls: getReturnValue().setSucceeded(true); and after that follows some logic, then if this logic fails, outcome of this command will be presented as successful. Compensation/rollbacks etc. are called, but data sent to user are wrong. example: if commands ends with: if (1 == 1) throw new RuntimeException(); getReturnValue().setSucceeded(true); rest will return: <fault> <reason>Operation Failed</reason> <detail>[Internal Engine Error]</detail> </fault> but if command would be like this: getReturnValue().setSucceeded(true); if (1 == 1) throw new RuntimeException(); result will be: <action> <status> <state>complete</state> </status> </action> How reproducible: 100% Actual results: see above. Expected results: command failed in both scenarios and returned data should reflect that.
Comment 1 Oved Ourfali 2014-06-12 06:27:30 UTC
I think that it should be the responsibility of the command to call setSucceeded(true) only once nothing is left to do... in my opinion, commands that do further logic after that assumes that this logic is insignificant for the successful completion of the command. Yair/Ravi - thoughts?
Comment 2 Martin Mucha 2014-06-12 07:35:45 UTC
I think, on contrary, that if somewhere there need not to be some convention(setSucceed must be last call of command) or faith in assumptions("insignificant" code will not fail), there should not be such things. Exception can occur unexpectedly, and so does db rollback. Flagging each command from which exception 'leaks' as unsuccessful will spell more robust architecture — user should not get info "OK" if it's not OK, and we should not rely on each command to behave as expected to provide user right response. It's command 'invoker' responsibility.
Comment 3 Yair Zaslavsky 2014-06-12 08:01:36 UTC
I agree with Martin. Bare in mind that execute is run from within CommandBase, so I think we should handle it in CommandBase.
Comment 4 Petr Beňas 2014-10-21 15:37:35 UTC
Comment 5 Eyal Edri 2015-02-17 17:07:32 UTC
rhev 3.5.0 was released. closing.