Wow, very sorry to see that. The last few weeks of threads have been increasingly worrisome. I read a very candid interview with Netflix CIO recently about AWS. He noted that AWS is always behind feature- wise because they never want to screw things up for their production users. My sense is that G. intends to gain share by an ever-increasing feature set, and the tradeoff is situations like these. Swords are always double-sided it seems. -stevep
On Apr 6, 7:49 pm, Ray <windz...@gmail.com> wrote: > The images are here: (Google Groups had problem on file attachment > apparently) > > http://dl.dropbox.com/u/182081/photo1.pnghttp://dl.dropbox.com/u/182081/photo2.pnghttp://dl.dropbox.com/u/182081/photo3.png > > > > > > > > On Saturday, April 7, 2012 10:47:52 AM UTC+8, Ray wrote: > > > AppEngine has had some pretty bad days recently. I was lucky enough to > > capture two dashboard with just a few hours apart which shows how bad the > > HR datastore could be: > > > The first two image show the hourly "Avg Latency" of the requests, the > > first one was the data around 8 hours ago, the other one was a few hours > > back. Normally the requests finish in a few hundreds ms, which are mostly > > get by key datastore access. When HR datastore misbehave, there are many > > deadline exceeded error right on db access and even if they complete, they > > become very slow. (from you chart you can see 3s to 9s latency in average). > > The two time periods are having similar traffic and no special code was > > running to cause the problem. > > > There were no code change, no abnormal status on the status page. They > > are typical (although this one is worst than normal in terms of error > > percentage) HR performance for me, HR datastore has constant errors, just > > not worst than M/S. > > > The third image shows how unstable HR datastore could be. Those errors are > > mostly caused by timeout on db access. > > > On Saturday, April 7, 2012 5:53:04 AM UTC+8, stevep wrote: > > >> Ray wrote: " I am still having around 1% percentage of datastore > >> timeout from any kind of datastore accesses like a simple > >> get_wih_id()." > > >> I have not seen this. If you and others are seeing 1% get_by_id()s > >> failing due to timeouts, it needs to be diagnosed by G. stat. Be sure > >> to report it. > > >> If it is a common issue, and remains so with G. acknowledging its part > >> and parcel of their PaaS, then it really is time to move to AWS for > >> me. > > >> -stevep > > >> On Apr 5, 6:26 pm, Ray <windz...@gmail.com> wrote: > >> > No one regret moving to HRD not because its great, instead its the M/S > >> > sucked too bad lately. 30 minutes download evey week is a disaster for > >> > any kind of app hosting on it. > > >> > At the mean time, HRD still isn't flawless. I am still having around > >> 1% > >> > percentage of datastore timeout from any kind of datastore accesses > >> like a > >> > simple get_wih_id(). It is slower than M/S and cost more (the /G cost > >> is > >> > now the same since Google raised the price for M/S too, but the slower > >> > respond still cost more on instance number) > > >> > The most stinky part is how Google advertise it as a 100% up time > >> service > >> > when it is generating more exceptions than most shared MySQL hosting > >> > outside. I know I know, its about scalability. But how many apps are > >> > having hundreds or thousands of DB access every second that can benefit > >> > from its scalability which can compensate the constant flaw? > > >> > On Thursday, April 5, 2012 4:46:57 AM UTC+8, Robert Kluin wrote: > > >> > > I think this is a really good idea. It will encourage people with > >> > > apps to move, and help further persuade new apps to not use > >> > > master-slave. I have not spoken to anyone who regretted the move to > >> > > high-replication. > > >> > > Robert > > >> > > On Wed, Apr 4, 2012 at 15:42, Chris Ramsdale <cramsd...@google.com> > >> wrote: > >> > > > Hey App Engine Users, > > >> > > > Almost 4 years after launch, we’ve seen fantastic growth both in > >> Google > >> > > App > >> > > > Engine and the apps that run on it. And although the Master/Slave > >> > > Datastore > >> > > > was a big part of our early success, it's time to announce the > >> > > deprecation > >> > > > of the Master/Slave Datastore in favor of the High-Replication > >> Datastore > >> > > > (HRD). HRD has provided us with higher availability and better, > >> more > >> > > > predictable performance. Many upcoming features will be HRD-only > >> and we > >> > > > strongly encourage you to migrate all your applications as soon as > >> > > possible > >> > > > using the migration tools found in the Application Settings tab of > >> the > >> > > > Administration Console. The deprecation period will follow the > >> guidelines > >> > > > set in our terms of service. If you have any questions or concerns, > >> > > please > >> > > > contact us at ms-datastore-deprecat...@google.com. > > >> > > > Chris > > >> > > > Product Manager, Google App Engine > > >> > > > -- > >> > > > 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. -- 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.