Hi all, In the graceful shutdown, we keep the member as active until it receives ReadyToShutdown event. In between these state changes, member will become unavailable. So the LB needs to aware of that, in order to avoid sending requests to a member which is going to graceful shutdown. I think that if we introduce another state say Maintenance in the Topology, before the actual shut down of the application inside the member takes place, then the LB will act upon the Maintenance state and remove the member from the active members list.
So the state changes of a member in the case of graceful shutdown would be: Created --> Starting --> Activated --> Maintenance --> ReadyToShutDown -->Terminated WDYT? Thanks, Reka On Mon, Jan 20, 2014 at 10:52 PM, Imesh Gunaratne <[email protected]> wrote: > Thanks Reka! I thought clean.sh is asynchronous. That's why I got that > question. What you have explained is great. > > Thanks > On Jan 19, 2014 11:52 PM, "Reka Thirunavukkarasu" <[email protected]> wrote: > >> Hi Imesh, >> >> On Sat, Jan 18, 2014 at 3:31 AM, Imesh Gunaratne <[email protected]>wrote: >> >>> Hi Reka, >>> >>> Great work! I have one question regarding the following step: >>> - Member will send the ReadyToShutDown event to instance-status topic >>> once it is done with the cleaning up >>> >>> Here I could see that we have introduced a new extension called clean.sh >>> to let the instance to execute a clean operation before the shutdown >>> happens. >>> My question is how does the instance communicate back to the cartridge >>> agent (Bash to Java) once its process is complete to send the >>> ReadyToShutDown event? >>> >> >> Yah. It is a good finding. I have handled it as the way where we handled >> to start servers via startServerExtension using start-server.sh and then >> publish Activated event. I believe that the executeCleanupExtension and the >> publishInstanceReadyToShutdownEvent are synchronous. So, after successfully >> executed the executeCleanupExtension, the method call returns and >> publishInstanceReadyToShutdownEvent will get executed. Please do correct >> me, if i'm wrong.. >> >> >> Thanks, >> Reka >> >>> >>> Thanks >>> Imesh >>> >>> >>> On Wed, Jan 15, 2014 at 1:48 PM, Lahiru Sandaruwan <[email protected]>wrote: >>> >>>> Hi Reka, >>>> >>>> Yes, I'm done. Sorry for the late reply. >>>> Following table explains it. >>>> >>>> >>>> Spawn invoked from AS >>>> >>>> Adding a pending member >>>> >>>> Member Activated event(from CC) >>>> >>>> Pending -> Active >>>> >>>> Graceful shutdown event sent from AS >>>> >>>> Active -> Termination pending >>>> >>>> Member Terminated event(from CC) >>>> >>>> Removing the termination pending member >>>> >>>> Member fault event(from CEP) >>>> >>>> Active Member -> Obsolete member >>>> >>>> I'm not sure if we can use ready shutdown state here. What autoscaler >>>> need to know is whether the member needs to be considered as a Active >>>> serving member or it is on the way to shut down(or already shut down). >>>> >>>> >>>> Thanks. >>>> >>>> >>>> On Wed, Jan 15, 2014 at 1:59 PM, Reka Thirunavukkarasu >>>> <[email protected]>wrote: >>>> >>>>> Hi Lahiru, >>>>> >>>>> Have you done with the changes on this improvement? If so, Can you >>>>> explain that how have you done with it? >>>>> Else, I believe that we can use the already existing ReadyToShutDown >>>>> state to improve the instance termination in autoscaler? >>>>> >>>>> Thanks, >>>>> Reka >>>>> >>>>> >>>>> On Tue, Jan 14, 2014 at 12:07 AM, Lahiru Sandaruwan >>>>> <[email protected]>wrote: >>>>> >>>>>> HI Reka, >>>>>> >>>>>> I think we have to do a improvement in autosclaer side since this >>>>>> will take some time to terminate. May be introducing "pending terminate". >>>>>> >>>>>> Otherwise scaling down rule wont stop scaling down properly... >>>>>> >>>>>> Thanks. >>>>>> >>>>>> >>>>>> On Fri, Jan 10, 2014 at 10:40 AM, Reka Thirunavukkarasu < >>>>>> [email protected]> wrote: >>>>>> >>>>>>> Hi Udara, >>>>>>> >>>>>>> +1. It is a good point. I will consider introducing a timeout in >>>>>>> Autoscaler to decide on the actual termination after making the call to >>>>>>> SM. >>>>>>> >>>>>>> Thanks, >>>>>>> Reka >>>>>>> >>>>>>> >>>>>>> On Fri, Jan 10, 2014 at 8:26 AM, Udara Liyanage <[email protected]>wrote: >>>>>>> >>>>>>>> Let's say Instance cleanup hang due to some reason(a file could not >>>>>>>> be deleted). Then instance won't send the readyToTerminate message. So >>>>>>>> isn't it better instance is terminated by force after timeout when the >>>>>>>> graceful shutdown is issued. >>>>>>>> >>>>>>>> Touched, not typed. Erroneous words are a feature, not a typo. >>>>>>>> >>>>>>> >>>>>>> >>>>>>> >>>>>>> -- >>>>>>> Reka Thirunavukkarasu >>>>>>> Software Engineer, >>>>>>> WSO2, Inc.:http://wso2.com, >>>>>>> Mobile: +94776442007 >>>>>>> >>>>>>> >>>>>>> >>>>>> >>>>>> >>>>>> -- >>>>>> -- >>>>>> Lahiru Sandaruwan >>>>>> Software Engineer, >>>>>> Platform Technologies, >>>>>> WSO2 Inc., http://wso2.com >>>>>> lean.enterprise.middleware >>>>>> >>>>>> email: [email protected] cell: (+94) 773 325 954 >>>>>> blog: http://lahiruwrites.blogspot.com/ >>>>>> twitter: http://twitter.com/lahirus >>>>>> linked-in: http://lk.linkedin.com/pub/lahiru-sandaruwan/16/153/146 >>>>>> >>>>>> >>>>> >>>>> >>>>> -- >>>>> Reka Thirunavukkarasu >>>>> Software Engineer, >>>>> WSO2, Inc.:http://wso2.com, >>>>> Mobile: +94776442007 >>>>> >>>>> >>>>> >>>> >>>> >>>> -- >>>> -- >>>> Lahiru Sandaruwan >>>> Software Engineer, >>>> Platform Technologies, >>>> WSO2 Inc., http://wso2.com >>>> lean.enterprise.middleware >>>> >>>> email: [email protected] cell: (+94) 773 325 954 >>>> blog: http://lahiruwrites.blogspot.com/ >>>> twitter: http://twitter.com/lahirus >>>> linked-in: http://lk.linkedin.com/pub/lahiru-sandaruwan/16/153/146 >>>> >>>> >>> >> >> >> -- >> Reka Thirunavukkarasu >> Software Engineer, >> WSO2, Inc.:http://wso2.com, >> Mobile: +94776442007 >> >> >> , >> > -- Reka Thirunavukkarasu Software Engineer, WSO2, Inc.:http://wso2.com, Mobile: +94776442007
