Your app actually only runs out of one data center at once. The diagram was a little misleading. If a data center goes down then your app is served out of the other with a different memcache (empty)
On May 14, 3:26 pm, Sergey Schetinin <ser...@maluke.com> wrote: > I'm pretty sure the memcached clusters (if there are more than one) > are not synchronized. First of all, that would be way too slow. > Second, the talk I referenced specifically mentions that when the apps > are being migrated from a DC, the memcache writes return success but > are in fact noop, because synchronizing memcache data does not make > sense. So I would expect that there's no synchronization going on > during regular operation as well. > > Anyway, I really hope that there's only one memcached cluster active at a > time. > > On 14 May 2011 22:05, rekby <timofey.koo...@gmail.com> wrote: > > > > > > > > > > > I thing HR-applications have more than one syncronised memcache - by > > me test save in HR-memcache in 3 times slower, than Master/Slave > > application. > > > On May 14, 6:56 pm, Sergey Schetinin <ser...@maluke.com> wrote: > >> So, I was watching the presentation on the HR datastore from the IO > >> 2011 > >> (http://www.google.com/events/io/2011/sessions/more-9s-please-under-th... > >> ) and one thing caught my attention: the slides were showing the > >> frontend instances running in more than one datacenter at the same > >> time. So I understand that the memcached can lose data at any time and > >> if the app is migrated between the datacenters all of the data in the > >> cache are lost, however, running the app in two or more datacenters at > >> a time each DC having a separate memcached cluster, that changes the > >> properties quite significantly. > > >> For example let's consider an app that uses some memcached key to keep > >> a cursor where to write to the datastore, atomically incrementing it > >> after each write. Such an app can detect if such a key is not present > >> in the cache and determine what the cursor is by doing a query on the > >> database and then add it to the cache atomically. If we add the > >> possibility that there's another DC running the app w/ an independent > >> memcached instance, such an app would just corrupt its own data. > > >> I hope I explained my concern well enough, and I would love to hear an > >> answer from someone on the App Engine team. > > >> Thank you. > > >> -Sergey > > >> --http://self.maluke.com/ > > > -- > > You received this message because you are subscribed to the Google Groups > > "Google App Engine" group. > > To post to this group, send email to google-appengine@googlegroups.com. > > To unsubscribe from this group, send email to > > google-appengine+unsubscr...@googlegroups.com. > > For more options, visit this group > > athttp://groups.google.com/group/google-appengine?hl=en. > > --http://self.maluke.com/ -- You received this message because you are subscribed to the Google Groups "Google App Engine" group. To post to this group, send email to google-appengine@googlegroups.com. To unsubscribe from this group, send email to google-appengine+unsubscr...@googlegroups.com. For more options, visit this group at http://groups.google.com/group/google-appengine?hl=en.