Yeah..it seems that way...but Imagine a scenario where there are a lot of writes... But few reads... Like...tracking user installs...and queries that occasionally need to see the number of installs... Unless that query is being run often... It will fall behind and anyone waiting for it to reindex won't be very happy. update after doesn't solve that ..so a HEAD every request every minute or so keeps it relatively up to date.... So you guarantee the client won't wait too long...Sven of they haven't made a read request in a while On Nov 20, 2013 3:01 PM, "Mark Hahn" <[email protected]> wrote:
> > .it sounds hackey. > > Just use update_after. It's not a hack. > > > On Wed, Nov 20, 2013 at 2:59 PM, Stanley Iriele <[email protected] > >wrote: > > > Idk..it sounds hackey.. But curl and crontab is good enough for me for > the > > views that can't fall more than 1 minute behind > > On Nov 20, 2013 2:57 PM, "Robert Newson" <[email protected]> wrote: > > > > > The bigcouch merge will not bring any automatic view updating > > > scheduler. Nothing stops someone contributing one, of course. > > > > > > B. > > > > > > > > > On 20 November 2013 22:49, Mike Marino <[email protected]> wrote: > > > > There are, of course, ways to get couchdb to update views dependent > on > > > > writes. I also believe this is supposed to get easier in the future > > > > (included in the bigcouch merge?). > > > > > > > >> Am 20.11.2013 um 23:46 schrieb Simon Metson <[email protected]>: > > > >> > > > >> Nope, views are updated on read, hence the "blocking" behaviour you > > > describe. You can query with update_after, which returns the stale > index > > > then triggers the update. > > > >> > > > >> > > > >>> On Wednesday, 20 November 2013 at 22:43, Mark Hahn wrote: > > > >>> > > > >>> I thought that every write triggered a view rebuild and that the > > stale > > > >>> option only meant a read didn't have to wait for a current rebuild > to > > > >>> finish. That would means the views are pretty much up-to-date. > > > >>> > > > >>> > > > >>>> On Wed, Nov 20, 2013 at 2:36 PM, Robert Newson < > [email protected]> > > > wrote: > > > >>>> > > > >>>> True, but remember couchdb doesn't automatically keep indexes > fresh > > in > > > >>>> the background, so "stale" can be "really really stale". ;) > > > >>>> > > > >>>> B. > > > >>>> > > > >>>> > > > >>>>> On 20 November 2013 22:34, Simon Metson <[email protected]> > > wrote: > > > >>>>> Unless your app can deal with querying the view stale. > > > >>>>> > > > >>>>> > > > >>>>>> On Wednesday, 20 November 2013 at 21:56, Mark Hahn wrote: > > > >>>>>> > > > >>>>>> I meant http view requests were blocked. It is waiting for the > > view > > > >>>>>> rebuild. > > > >>>>>> > > > >>>>>> I'm can't type what I'm thinking today. > > > >>>>>> > > > >>>>>> > > > >>>>>>> On Wed, Nov 20, 2013 at 1:54 PM, Mark Hahn <[email protected]> > > > wrote: > > > >>>>>>> > > > >>>>>>> never mind. I wasn't talking about the file level at all. I > meant > > > that > > > >>>>>>> http read requests are blocked after http update requests. > > > >>>>>>> > > > >>>>>>> > > > >>>>>>> On Wed, Nov 20, 2013 at 1:52 PM, Robert Newson < > > [email protected] > > > > > > > >>>> wrote: > > > >>>>>>> > > > >>>>>>>> "DB reads are blocked by DB updates at the http level." > > > >>>>>>>> > > > >>>>>>>> Nope, there's a process that can read the database and a > > separate > > > >>>> one > > > >>>>>>>> for writing to it. Writing to an append only file is obviously > > > >>>>>>>> serialized but there's no need to block reads. > > > >>>>>>>> > > > >>>>>>>> B. > > > >>>>>>>> > > > >>>>>>>> > > > >>>>>>>> On 20 November 2013 21:35, Mark Hahn <[email protected]> > wrote: > > > >>>>>>>>>> Database writes are not coupled to view updates. > > > >>>>>>>>> > > > >>>>>>>>> > > > >>>>>>>>> > > > >>>>>>>>> I understand now, you are talking about file read/write > level. > > DB > > > >>>> reads > > > >>>>>>>>> are blocked by DB updates at the http level. > > > >>>>>>>>> > > > >>>>>>>>> > > > >>>>>>>>> On Wed, Nov 20, 2013 at 1:05 PM, Robert Newson < > > > >>>> [email protected] > > > >>>>>>>>> wrote: > > > >>>>>>>>> > > > >>>>>>>>>> "A write requires updating views and reads have to wait for > > the > > > >>>> update" > > > >>>>>>>>>> > > > >>>>>>>>>> Is not true. Database writes are not coupled to view > updates. > > > >>>>>>>>>> > > > >>>>>>>>>> Sent from my iPad > > > >>>>>>>>>> > > > >>>>>>>>>>> On 20 Nov 2013, at 20:59, Mark Hahn <[email protected]> > > wrote: > > > >>>>>>>>>>> > > > >>>>>>>>>>> A write requires updating views and reads have > > > >>>>>>>>>>> to wait for the update > > > >>>>>>>>>>> > > > >>>>>>>>>> > > > >>>>>>>>>> > > > >>>>>>>>> > > > >>>>>>>> > > > >>>>>>> > > > >>>>>> > > > >>>>> > > > >>>> > > > >>>> > > > >>> > > > >>> > > > >>> > > > >> > > > >> > > > > > >
