If you see something like the following message in the body of a log warning entry, then that request has been counted as a "high CPU request"
This request used a high amount of CPU, and was roughly X times over the average request CPU limit. High CPU requests have a small quota, and if you exceed this quota, your app will be temporarily disabled. A request is counted as a "High CPU Request" if it uses more than one thousand runtime CPU megacycles. For now, your app receives 2 "High CPU Request Credits" per minute, and can store up to 60 of these credits at any one time. In the future, we plan to reduce the impact of high CPU requests on the server, but for now you'll need to optimize expensive portions of your app in order to maintain throughput for the App Engine system as a whole. I'll post an FAQ on high CPU requests soon. Thank you, Jeff On Dec 2, 9:09 am, "[EMAIL PROTECTED]" <[EMAIL PROTECTED]> wrote: > PS All my datastore operations are queries (no puts). And there are > about 1500 records in all. > > On Dec 2, 5:03 pm, "[EMAIL PROTECTED]" <[EMAIL PROTECTED]> > wrote: > > > > > I have been playing with appengine for a couple of weeks and it is a > > blast! Thanks so much. > > > It is also nice to see that people have already asked the exact > > question on one's mind. > > > I have a situation identical to what Waldemar Kornewald descibes. > > > What I cannot fathom from the responses (I apologize in advance if > > this is obvious to everyone else) in this thread is whether and when > > theseCPU-intensive datastore warnings are a problem to be concerned > > about if the totalCPUquota is not exceeded. > > > Best Regards > > --RG > > > On Nov 12, 12:57 am, Jeff S <[EMAIL PROTECTED]> wrote: > > > > Apologies for the late response Josh. To clarify, when looking at the > > > logs, the coloredwarningmessage with the exclamation-mark icon on > > > the collapsed log entry is based on both the runtime and datastoreCPU > > > combined. ACPUmessage within the expanded log entry measures the > > > runtimeCPUonly. > > > > To reuse Waldemar's screenshot, the circledwarningmessage is the > > > combined datastore and runtimeCPU: > > > >http://freenet-homepage.de/wkornewald/logs-expanded.jpg > > > > Happy coding, > > > > Jeff > > > > On Oct 29, 6:46 pm, Josh Heitzman <[EMAIL PROTECTED]> wrote: > > > > > Jeffyou said: > > > > > "TheCPUwarning in the log is based on runtimeCPU, while the > > > > displayedCPUwarning in the log is based on runtime and datastoreCPU." > > > > > but you wrote "in the log is based on" both times. I'd assume one of > > > > those is type, so could you clarify which if the mcycle consumption > > > > value shown in the per request logs the runtime mcycles or the runtime > > > > and datastore mcycles? > > > > > If the mcycle value shown in the per request logs is only the runtime > > > > mcycles then issue 786 is not a duplicate of 814 as the values listed > > > > in 786 were taken from the per request logs. > > > > > Thanks, > > > > > Josh Heitzman > > > > > On Oct 29, 11:27 am,JeffS <[EMAIL PROTECTED]> wrote: > > > > > > Hi Waldemar, > > > > > > On Oct 28, 3:22 pm, Waldemar Kornewald <[EMAIL PROTECTED]> wrote: > > > > > > > HiJeff, > > > > > > > On 28 Okt., 21:11,JeffS <[EMAIL PROTECTED]> wrote: > > > > > > > > > Our queries and indices are rather simple. I should clarify: > > > > > > > > The particular problem I'm talking about is a view that > > > > > > > > currently has > > > > > > > > a response time of about 360ms and almost all of the time > > > > > > > > (280ms) is > > > > > > > > spent on datastore put()/fetch()/get(). IOW, runtimeCPUusage is > > > > > > > > low. > > > > > > > > If you only read data then 360ms would probably consume less > > > > > > > > than > > > > > > > > 1500mcycles, but since this view primarily writes to the > > > > > > > > datastore it > > > > > > > > consumes more than 10,000mcycles! That's really bothering me > > > > > > > > because I > > > > > > > > already get yellow warnings for views that consume more than > > > > > > > > 1000mcycles and here I have red warnings. How am I supposed to > > > > > > > > implement a simple write operation if it eats mcycles so > > > > > > > > quickly? Can > > > > > > > > I just ignore thewarningsince only runtimecpucounts and we > > > > > > > > primarily consume datastorecpuin this case? > > > > > > > > From your description, it sounds like the datastoreCPUusage need > > > > > > > not > > > > > > > concern you. If theCPUUsage listed on the admin console is at an > > > > > > > acceptable level and the logs do not contain large amounts > > > > > > > ofhighCPU > > > > > > > request warnings, then your app should be okay. > > > > > > > Well, that's the problem. The dashboard shows awarningsign ("<!>") > > > > > > and veryhigh"AvgCPU" usage for that URL, but the logs don't mention > > > > > > anything at all for that URL. So, which is more correct? The > > > > > > dashboard > > > > > > or the logs? :) > > > > > > The two are currently measuring different things. TheCPUwarning in > > > > > the log is based on runtimeCPU, while the displayedCPUwarning in > > > > > the log is based on runtime and datastoreCPU. > > > > > > Thank you, > > > > > >Jeff > > > > > > > Normally, if I consume too muchCPUI get an explicitwarning > > > > > > *message* in the logs: > > > > > > > 10-25 06:48AM 43.525 / 200 802ms 2754mcycles <!> 3kb > > > > > > [...snip...] > > > > > > This request used ahighamount ofCPU, and was roughly 2.6 > > > > > > times over the average requestCPUlimit.HighCPUrequests have a > > > > > > small quota, and if you exceed this quota, your app will be > > > > > > temporarily disabled. > > > > > > > As I said, thatwarningmessage doesn't appear. But if I print > > > > > > something to the logs I can see in the header of the log entry that > > > > > > the request consumes a huge amount ofCPU(e.g., "15000mcycles <!>", > > > > > > please note the "<!>"warning*sign*). Why do I *not* get awarning > > > > > > *message* although I can provoke awarning*sign* if (and only if) I > > > > > > print something to the logs? > > > > > > > Also, why does theCPUquota usage go up and down so quickly? A few > > > > > > minutes ago I had 40gcycles used and after I visited thathigh-CPUURL > > > > > > four times it went up to 85gcycles and a few minutes later it went > > > > > > back to 0. I thought this was a *24h* moving window? Don't get me > > > > > > wrong. I don't mind quota usage going back to zero every few minutes > > > > > > (or seconds would be fine, too). :) > > > > > > > Thanks! > > > > > > > Bye, > > > > > > Waldemar Kornewald --~--~---------~--~----~------------~-------~--~----~ 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 [EMAIL PROTECTED] For more options, visit this group at http://groups.google.com/group/google-appengine?hl=en -~----------~----~----~----~------~----~------~--~---