Today, my app has seen 2 timeout errors at 3:10 pm according to the log. Each one was 30 seconds apart. One was a Get and the other a Put. Each on totally different classes that are not connected in any way.
On Apr 17, 4:15 pm, Brandon Thomson <gra...@gmail.com> wrote: > Google will not really acknowledge this as a problem or defect but > neither do they provide a lot of options for workaround. My experience > is that Timeouts do occur on about 0.5% of puts regardless of size of > entity. These are root entities. It is not caused by write contention > from multiple requests on the entity. A handler only called by > periodic cron job will experience it. > > For one week in March when they were playing around with the > architecture Timeouts went from about 0.5% of all puts to about 5%. > Since it has been not as bad. There was an open defect but they > recently closed > it:http://code.google.com/p/googleappengine/issues/detail?id=764 > > I agree with notcourage about queues being a good solution and this is > what I do because the data has to go somewhere. Currently I am using > SQS. But a perhaps more important question is, is the time you spend > implementing these workarounds better spent implementing your app on a > different platform? I honestly don't know the answer, maybe notcourage > has a better idea. I am curious to hear any more thoughts you have. > Because if we are trying to make money this is always at front of > mind. > > On Apr 17, 12:14 pm, notcourage <klr...@gmail.com> wrote: > > > Good for you that your company has good prospects. As somebody who has > > built high traffic sites (co-founder of IGN.com), I offer some advice: > > Don't even think about driving traffic to it unless it has been stable > > for weeks. If you are having problems w/ minute traffic, you can't > > imagine how bad it will be under load. Plan for graceful degradation. > > Some possible degradations: Read-only mode except for existing > > registered users. Static site. > > > Another thing to try is to queue put's in memcache, in a simple > > datastore q, or inhttp://aws.amazon.com/sqs/. The deferred put's can > > be stored as simple Strings using the Pickler. > > > On Apr 17, 5:12 am,RayMalone <rayish...@gmail.com> wrote: > > > > Speaking only for my app, I can tell this is not an application > > > issue. My datastore is made up of only root entities and in most > > > cases works well quickly. My app doesn't see large volumes of > > > requests per second and it's more like requests per minute. There > > > is no possible way the errors I'm seeing are from contention given the > > > minutes between requests. And, the same exact process is handled a > > > few hours before within 600ms. My app is soon to increase in > > > volume and needs to be stable. I love using app engine, but I've > > > never faced errors like this on any other platform from IIS Server and > > > SQL Server to PHP and MY SQL or even Unix and Oracle. Don't get me > > > wrong, I understand the difference in the platforms, If there is > > > anything I can do to prevent this I would love to know. I'm not > > > moving my app, but need to find a solution. My company will be on the > > > local NBC news (I will be mentioning app engine) and may see a huge > > > volume hitting the site next week. > > > > On Apr 17, 4:54 am, Sylvain <sylvain.viv...@gmail.com> wrote: > > > > > Datastoretimeoutis one of the biggest (and oldest) issue with GAE > > > > and mostly because it is random. > > > > You can get it on get, put, fetch,... even with few entities. > > > > > I hope that soon, datastoretimeoutwill be negligible. Currently, it > > > > is not the case. > > > > > Regards > > > > > On 17 avr, 09:50, Paul Kinlan <paul.kin...@gmail.com> wrote: > > > > > > Hi, > > > > > > I would just like to add myself into this, my app twitterautofollow > > > > > (twollo) > > > > > regularly gets DataStore time outs on puts (in the most part). All my > > > > > entities are root entities. > > > > > > I did have a thread open on this only a few days ago. > > > > > > Paul > > > > > > 2009/4/17RayMalone <rayish...@gmail.com> > > > > > > > I'm seeing random timeouts where there are many seconds or minutes > > > > > > between requests and hours before app engine handled the same > > > > > > request > > > > > > within 400ms. I hit another one today that will actually cost my > > > > > > business, not much but some. My app needs to catch these and I am > > > > > > on > > > > > > many pain points. However, I think the timeouts needs addressed. > > > > > > > On Apr 16, 7:10 pm, DarkCoiote <darkcoi...@gmail.com> wrote: > > > > > > > Yes... I've read that... > > > > > > > > The problem with using all entities as root is that I'm unable to > > > > > > > use > > > > > > > transactions as it is... I would have to code a lit bit... > > > > > > > although > > > > > > > I just found a paper describing a project that seems really good. > > > > > > > >http://danielwilkerson.com/dist-trans-gae.html > > > > > > > > I think it will be presented this weekend, and I really hope that > > > > > > > it goes "public" or even yet plugged to app engine! > > > > > > > > Just double-checked my GAE logs, and it seems that I'm getting > > > > > > > timeouts in "get" operations as well... multiple requests could > > > > > > > cause > > > > > > > that too (one get 'over' a put for example)... > > > > > > > > Have to check my code.... > > > > > > > > Thank you > > > > > > > > On Apr 16, 2:10 pm, "Jeff S (Google)" <j...@google.com> wrote: > > > > > > > > > On Thu, Apr 16, 2009 at 4:16 AM, DarkCoiote > > > > > > > > <darkcoi...@gmail.com> > > > > > > wrote: > > > > > > > > > > On Apr 15, 3:26 pm, "Jeff S (Google)" <j...@google.com> wrote: > > > > > > > > > > HiRay, > > > > > > > > > > > Which operation was is that timed out (get, query, put)? > > > > > > > > > > Also, how > > > > > > > > > > consistently are you seeing these timeouts? > > > > > > > > > > > I generally recommend catching datastore timeouts and > > > > > > > > > > handling them > > > > > > in a > > > > > > > > > way > > > > > > > > > > that makes sense for your app. There are currently > > > > > > > > > > occasional > > > > > > (quite rare > > > > > > > > > as > > > > > > > > > > a percentage) timeouts for queries and gets, and timeouts > > > > > > > > > > > on a put is often > > > > > > > > > > an indicator of contention on that entity or entity group. > > > > > > > > > > well... almost (if not all) of the timeouts I'm seeing are on > > > > > > > > > 'put' > > > > > > > > > operations... > > > > > > > > > but contention would need, like, 2 or more operations on the > > > > > > > > > same > > > > > > > > > entity > > > > > > > > > (all my entities are roots... stupid, I know ), right? > > > > > > > > > Actually, making most of your entities roots is often better > > > > > > > > than deep > > > > > > > > ancestor trees in terms of overall write throughput. In a > > > > > > > > transactional > > > > > > > > write a entity is updated which has a parent entity, the > > > > > > > > ancestors are > > > > > > > > locked. So if nearly concurrent requests update different > > > > > > > > entities > > > > > > which all > > > > > > > > share a common ancestor (in other words the entities are in the > > > > > > > > same > > > > > > entity > > > > > > > > group), some of the child entity updates could fail due to > > > > > > > > contention > > > > > > on > > > > > > > > writes. There are more details here in the documentation: > > > > > > > > >http://code.google.com/appengine/docs/python/datastore/keysandentityg. > > > > > > .. > > > > > > > > > """ > > > > > > > > Tips for using entity groups: > > > > > > > > > - Only use entity groups when they are needed for transactions. > > > > > > > > For > > > > > > other > > > > > > > > relationships between entities, use ReferenceProperty > > > > > > > > properties and > > > > > > Key > > > > > > > > values, which can be used in queries. > > > > > > > > > - The more entity groups your application has—that is, the more > > > > > > > > root > > > > > > > > entities there are—the more efficiently the datastore can > > > > > > > > distribute > > > > > > the > > > > > > > > entity groups across datastore nodes. Better distribution > > > > > > > > improves the > > > > > > > > performance of creating and updating data. Also, multiple users > > > > > > attempting > > > > > > > > to update entities in the same entity group at the same time > > > > > > > > will cause > > > > > > some > > > > > > > > users to retry their transactions, possibly causing some to > > > > > > > > fail to > > > > > > commit > > > > > > > > changes. Do not put all of the application's entities under one > > > > > > > > root. > > > > > > > > > - A good rule of thumb for entity groups is that they should be > > > > > > > > about > > > > > > the > > > > > > > > size of a single user's worth of data or smaller. > > > > > > > > > - Entity groups do not have a significant impact on the speed of > > > > > > queries. > > > > > > > > """ > > > > > > > > > If your entities are roots, then yes contention could occur if > > > > > > > > the same > > > > > > > > entity were updated by overlapping requests. > > > > > > > > > Thank you, > > > > > > > > > Jeff > > > > > > > > > > I'll check for bugs that could cause multiples requests and > > > > > > > > > stuff > > > > > > like > > > > > > > > > that.... > > > > > > > > > > Thank you > > > > > > > > > > > Happy coding, > > > > > > > > > > > Jeff > > > > > > > > > > > On Mon, Apr 13, 2009 at 4:11 PM, DarkCoiote > > > > > > > > > > <darkcoi...@gmail.com> > > > > > > > > > wrote: > > > > > > > > > > > > Looks like my and others problem... posted a few days ago: > > > > > > >http://groups.google.com/group/google-appengine/browse_thread/thread/. > > > > > > > > > .. > > > > > > > > > > > > Random datastore timeouts in totally unexpected places... > > > > > > > > > > > > On Apr 13, 2:36 pm,RayMalone <rayish...@gmail.com> wrote: > > > > > > > > > > > > I'm not sure why this occurs all of a sudden. The same > > > > > > > > > > > > job > > > > > > runs in > > > > > > > > > > > > 600ms. Then once in a while this runs over 6800 MS > > > > > > > > > > > > and times > > > > > > out. > > > > > > > > > > > > I can't have jobs justtimeoutfor no reason. > > > > > > > > > > > > According to > > > > > > my logs > > > > > > > > > > > > during this time the only other request was a second > > > > > > > > > > > > prior. > > > > > > This > > > > > > > > > > > > occurred at 04-13 09:13AM 29.627 today and other times > > > > > > > > > > > > at > > > > > > random in > > > > > > > > > > > > the last few days. --~--~---------~--~----~------------~-------~--~----~ 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 -~----------~----~----~----~------~----~------~--~---
