On 2015-02-25, 9:40 AM, "OC" wrote: Chuck,
On 25. 2. 2015, at 18:18, Chuck Hill <[email protected]<mailto:[email protected]>> wrote: I just thought of something else. Is this app running with a single instance? Yes, it is. If so, this might mean that the response was no received from the app before the Receive Timeout expired (see Adaptor Settings in on the application page in Java Monitor). Can it lead to this weird report? I always thought it leads to "No instance" pretty consistently. At least, it did when my app was really slow, before I moved the imports to a background task :) Yes, that might lead to this weird report. I have few apps that run single instance and those that do, don’t have any long running requests. So I am just guessing. There are three timeouts: Connect, Send, and Receive (response). IIRC, No Instance comes from the Connect timeout. This error message corresponds to something going wrong with receiving the response after Connect and Send (request) have succeeded. Chuck The solution to that is to make your app faster :-) or increase the timeout. Anyway, App.awake to App.sleep the request took 9.952 seconds -- it is a lot, but it should be safely below the default 30 s (and if the server admin changed that, he would hardly decrease the default). Is this app running without concurrent request dispatch enabled? Nope, it runs with concurrent dispatch enabled (which causes lot of grief by itself, for 70-odd per cent of the codebase is years old and has been written for serialized requests... but that's another story). ... That also explains why the app log does not correspond to the user’s perception. Most time, it does. Far as I can say, yesterday was the first time this happened (some cases might well be unreported of course). It sounds like either the CSV import is doing too much before handing off to the second stack, the second stack is not getting used correctly (I doubt, but will mention), or some other action in the app is slow. If there is a lot of constant database access (like a big save or big fetch) from one request, it will cause other requests (using the same OSC) to queue up for the EOF lock. That was the problem we hopefully cured lately in the thread “looong saveChanges in a background task”; but it did not happen at the moment -- far as I can say. (Well at least I still do log all the DB operations, and none were logged there.) Thanks again, OC On 2015-02-25, 8:57 AM, "OC" wrote: Oh, and since I'm writing anyway... On 24. 2. 2015, at 18:48, Chuck Hill <[email protected]<mailto:[email protected]>> wrote: I don’t see that one very often. IIRC, it means that it sent the request to an instance and got a null response back. That probably means the app threw an exception either very early or very late in the R-R loop. Thanks! Check the app logs for exceptions. None. Far as I can say, the request -- it was CSV import, too, but none of those which went wrong -- did spawn its background thread and finished all right -- I'm logging R/Rs from application's awake/sleep, and all looks OK. Definitely no exception in my application log nearby. Did an instance run out of memory? Hardly. I log the heap (and lately also the Perm Gen stuff which bit my back a couple of days ago), and there was plenty of both. Check the Apache error.log just to be sure. Thanks again, I've asked the system admin to do that. I did that anyway, for that day there _was_ something fishy (hmmmm... perhaps it even might be related): The thing is, I've got a report from a user that one request took more than 20 seconds to reload in his browser. I have found it in my logs, and App.awake to App.sleep it took 1.818 seconds. Another request reported to last a dozen of seconds from the user's point of view took 0.577 seconds App.awake to App.sleep. Weird. Thanks again, OC On 2015-02-24, 8:35 AM, "OC" wrote: Hello there, what the $subject means and who reports it? I've googled a bit before coming here, and found (a) it seems to be a Wonder error report (https://github.com/wocommunity/wonder/blob/master/Utilities/Adaptors/Adaptor/errors.h) (b) it seems to be quite common to see with Apple own applications (http://forums.appleinsider.com/t/418/apple-store-seems-to-be-down-right-now/40 and more similar ones, with Knowlegde Base, with Apple WebMail, etc.) (Incidentally, Apple uses Wonder? That's nice to know.) But what I haven't been able to find (yet) is what it actually means? :-O I did succeed to find the appopriate source (transaction.c in the Adaptor) and function (_errorResponse) which seem to generate the report, but without a detailed study of the adaptor code I'm none the smarter of what condition causes this (well, the condition is that WOAppReq.error contains 4, but... :)) Does somebody here know, or am I to dive into the adaptor sources? Not that I would care, but one of the users of my application did see the thing today :( Thanks a lot, OC _______________________________________________ Do not post admin requests to the list. They will be ignored. Webobjects-dev mailing list ([email protected]<mailto:[email protected]>) Help/Unsubscribe/Update your Subscription: https://lists.apple.com/mailman/options/webobjects-dev/chill%40gevityinc.com This email sent to [email protected]<mailto:[email protected]>
_______________________________________________ Do not post admin requests to the list. They will be ignored. Webobjects-dev mailing list ([email protected]) Help/Unsubscribe/Update your Subscription: https://lists.apple.com/mailman/options/webobjects-dev/archive%40mail-archive.com This email sent to [email protected]
