Thanks, thanks that answers my questions. William
On Wed, Jun 27, 2018, 5:45 PM VENKATESH KUMAR, VIJAY <[email protected]> wrote: > HI William, Yes, that queue is maintained in memory currently. If the > queue builds up (due to prolonged downstream issue), then there is risk of > event loss during restart. Typically though the queue count is maintained > at zero. > > > > Thanks, > > Vijay > > > > *From:* William Kurkian <[email protected]> > *Sent:* Wednesday, June 27, 2018 5:36 PM > *To:* VENKATESH KUMAR, VIJAY <[email protected]> > *Cc:* [email protected] > *Subject:* Re: [onap-discuss] DCAE event Guaranteed Delivery > > > > Thanks, that is very clear and helpful. > > > > I just have one more question. In the case where the messages are > accumulating on the queue in the VES, is that in memory? So if the memory > state is lost, then the messages are lost? > > > > I will be running this in an environment without full production support, > so I figuring out some failure cases. > > > > Thanks, > > William > > > > On Wed, Jun 27, 2018, 5:22 PM VENKATESH KUMAR, VIJAY <[email protected]> > wrote: > > Hi, > > Just to clarify, the acknowledgement from VESCollector primarily indicates > the message was received and validated in collector but not status of > complete processing (by collector/DCAE or rest of ONAP components for that > matter). > > > > As VES entails support for multiple domains/topics and batch mode, the > response to VNF (or PNF) was strictly decoupled from rest of collector > processing by design. We have also seen instances of DMAAP connection being > lost temporarily (either network or dmaap being down) during which > collector continues to receive events and queue them. Once connection is > restored, VESCollector will forward them to appropriate topics. By handling > these transient issue, we are avoiding impact on xNF’s (otherwise xNF’s > will need queue and resend events). On typical production setup, the FS > will be monitored and should be caught and addressed before > application/pods are crashing. DMaap being critical component in ONAP, > expectation is it will be highly-available and guarantee delivery for > subscribers (once messages are published into MR topic by collector) > > > > Hope this clarifies. > > > > Regards, > > Vijay > > > > > > *From:* [email protected] <[email protected]> *On > Behalf Of *William Kurkian > *Sent:* Wednesday, June 27, 2018 2:23 PM > *To:* [email protected] > *Subject:* [onap-discuss] DCAE event Guaranteed Delivery > > > > Hi, > > > > I've been experimenting with having a PNF send event information to the > DCAE, the VES specifically, in order to get it on the DMaaP. I've noticed a > failure scenario when the machine runs out of disk. A number of pods start > failing, including the DMaaP message-router and kafka. > > > > The problem I see is that, if the VES Pod is still running, it returns a > 200 Ok to any messages I send it, despite the DMaaP not running. > > > > This worries me because in a production system, faults could occur that > prevent messages from being deliverable, and certain events are > critical(alarms) and need to be guaranteed. Could someone help me > understand what is going on here? > > > > From what I see, the VES has a queue it puts the messages are on, but from > that point I don't see that I can depend on that for messages to survive in > case of things like reboot. > > > > Thanks, > > William Kurkian > > > > -=-=-=-=-=-=-=-=-=-=-=- Links: You receive all messages sent to this group. View/Reply Online (#10658): https://lists.onap.org/g/onap-discuss/message/10658 Mute This Topic: https://lists.onap.org/mt/22755059/21656 Group Owner: [email protected] Unsubscribe: https://lists.onap.org/g/onap-discuss/unsub [[email protected]] -=-=-=-=-=-=-=-=-=-=-=-
