UIMA-AS supports timeouts for remote annotators; the default timeout is infinite. On timeout uima-as will take the action specified by error handling configuration, but in any case the CAS sent to the remote will be available for reuse.
Eddie On Thu, Aug 25, 2022 at 12:05 PM Timo Boehme <timo.boe...@ontochem.com> wrote: > PS: if the OS is killing the process because of lack of memory (typical > case) it means the Java VM is allowed to used more (heap) memory as is > available at this node. Maybe consider to adjust the memory setting for > the Java process to prevent the OS kill. Then you may get an > OutOfMemoryException which is bad too but the JavaVM might be able to do > some cleanup/shutdown etc. > > > Timo Boehme > > > > Am 25.08.22 um 17:55 schrieb Timo Boehme: > > Hi Daniel, > > > > not using UIMA-AS myself but if the OS is killing a process because of > > lack of resources it normally does so with a hard kill which does not > > allow the Java VM process to do any shutdown work. > > One would need a separate process controlling the Java one and react if > > the Java VM is killed - however this won't help in getting the CAS > > released (or the controlling process has specific UIMA knowledge). I > > don't known if the UIMA-AS uses such a 2-process model per node but > > assume it does not. > > > > > > Regards, > > Timo Boehme > > > > > > Am 25.08.22 um 17:28 schrieb Daniel Cosio: > >> Yes, this is uima-as-jms.. the pipeline gets a signal from the os and > >> shutdown.. but we loose the CAS. Is there any api I can use to tell > >> the collection reader to invalidate. I know the AE has > >> A temp queue connection that communicates the CAS releases..I was > >> wonder if there was any way of getting the temp queue connection and > >> sending the message back to return the CAS.. Possible in a shutdown > hook. > >> > >> > >> Daniel Cosio > >> dcco...@gmail.com > >> > >> > >> > >>> On Aug 25, 2022, at 9:20 AM, Eddie Epstein <eaepst...@gmail.com> > wrote: > >>> > >>> Daniel, is this again a uima-as deployment? If so, since the OS kills > >>> processes, is it some remote AE being killed? > >>> > >>> Eddie > >>> > >>> On Wed, Aug 24, 2022 at 10:04 AM Daniel Cosio <dcco...@gmail.com> > wrote: > >>> > >>>> Hi, I have some instances where the OS has killed a pipeline to > recover > >>>> resources.. When this happens the pipeline never returns the CAS to > the > >>>> reader so the reader now has 1 less CASes > >>>> Available.. Is there a was to either > >>>> 1. Add a shutdown hook on the pipeline to return the CAS if it gets a > >>>> shutdown signal > >>>> or > >>>> 2. Set an expiration on the collection reader to expire a CAS that > >>>> is not > >>>> returns and issue a new one into the CAS pool > >>>> > >>>> > >>>> Thanks > >>>> > >>>> > >>>> Daniel Cosio > >>>> dcco...@gmail.com > >>>> > >>>> > >>>> > >>>> > >> > > > > > > > -- > OntoChem GmbH > Blücherstraße 24 > 06120 Halle (Saale) > Germany > > email: timo.boe...@ontochem.com | web: www.ontochem.com > HRB 215461 Amtsgericht Stendal | USt-IdNr.: DE246232735 > managing directors: Dr. Lutz Weber (CEO), Dr. Felix Berthelmann (COO) > >
