John Snow <js...@redhat.com> writes: > On 11/20/20 5:25 AM, Markus Armbruster wrote: >> John Snow <js...@redhat.com> writes: >>> On 11/11/20 3:27 AM, Markus Armbruster wrote: >>>> John Snow <js...@redhat.com> writes: >>>>> On 11/10/20 1:22 AM, Markus Armbruster wrote: >>>>>> John Snow <js...@redhat.com> writes: > >> >> If you find yourself in a situation where an error class would >> definitely be useful, we should talk about providing you one. >> > > OK, thanks. I just wanted to check there wasn't a stronger opposition > than I was aware of. I only vaguely knew they were "on the way out." > >>>> QGA has a special case, but you probably don't want to know. >>> >>> I do want to know, unfortunately. A good QMP client should likely >>> support both targets. >> >> QGA has a few commands that send *nothing* on success. >> >> qapi-code-gen.txt explains: >> >> Normally, the QAPI schema is used to describe synchronous exchanges, >> where a response is expected. But in some cases, the action of a >> command is expected to change state in a way that a successful >> response is not possible (although the command will still return an >> error object on failure). When a successful reply is not possible, >> the command definition includes the optional member 'success-response' >> with boolean value false. So far, only QGA makes use of this member. >> >> qmp-spec.txt neglects to cover this special case. >> > > Oh, I see. That's fun! > > (Was it not possible to have the client send an ACK response that > doesn't indicate success, but just signals receipt? Semantically, it > would be the difference between do-x and start-x as a command.)
Feels quite possible to me. If I read git-log correctly, the commands' design ignored the race with shutdown / suspend, and 'success-response': false was a quick fix. I think changing the commands from 'do-x' to 'start-x' would have been better. A bit more work, though, and back then there weren't any 'start-x' examples to follow, I guess. I wish success-response didn't exist, but is getting rid of it worth changing the interface? I honestly don't know. >>>> Even though counting responses is a workable way to map responses back >>>> to requests, I recommend to either have at most one request in flight >>>> (so no counting is necessary), or to add IDs to all requests (so >>>> counting isn't necessary as long as you get the syntax right enough to >>>> avoid the ID-less errors discussed above). >>> >>> I am taking this approach. I still need some error handling for the >>> case that an incoming response cannot be routed to the correct pending >>> queue. >>> >>> The QMP spec states that if I get an ID and I don't recognize it, I am >>> free to drop the response on the floor (So I have done so), >> >> Should not happen. When it does, something is wrong, possibly seriously >> wrong. > > Yep, I agree. > > (In my client library: I have decided to enforce string IDs and not > allow the caller to specify their own. Though we allow arbitrary objects > to be used as an ID, being able to compare more complex objects seems > more prone to failure than a simple string.) Makes sense. >> One possible scenario: ID got corrupted on the way from client via >> server back to the client. The response with the uncorrupted ID will >> never come. Other responses and events may still arrive fine. >> >> Two recovery strategies: >> >> 1. Drop the unexpected response, carry on. If commands are in flight, >> be prepared for missing responses. >> >> 2. Give up, close connection. >> >> Pick the strategy that best fits your client's needs. > > Dropping it on the floor seems fine for now, the spec says I can! :) > > The most realistic scenario for this outside of catastrophic error is > that a client sends a command, but then times out waiting for a > response. Later, the response arrives -- but that context has already > come and gone, so there's nowhere to route the message to. > > Effectively, an orphaned reply. I suppose I could always stash them in > an orphaned queue and a very, very rigorous client could check the > orphaned queue if something winds up in there, but I suspect most > clients would actually be just as happy to let it fall on the floor. Sounds okay to me. >>> but I need >>> to handle the case that I see a response with no ID at all. >> >> Should not happen as long as the client sends syntactically sane input. >> When it does, something is wrong, possibly seriously wrong. >> >>> Perhaps the easiest, and most sensible thing to do, is to declare this >>> a catastrophic failure and move the QMP connection into an error >>> state, invalidate the whole queue, and disconnect. >>> >>> (And then maybe some of the commands you issued got processed, maybe >>> they didn't. I guess it's up to the application driving the library to >>> query around to find out what happened.) >> >> Again, two recovery strategies: >> >> 1. Drop the unexpected response, carry on. If commands are in flight, >> be prepared for missing responses. The server's JSON parser may now >> be in a state where it can't parse additional commands. You may want >> to provoke a lexical error to force it into known-good state, as >> explained below. >> >> 2. Give up, close connection. >> > > 2 is easier for now. 1 involves some more advanced recovery logic, but > as you say: something is wrong, possibly seriously wrong. Taking the easy way out sounds okay to me. Build more advanced recovery when you *know* you need it. >>> I see. So OOB commands don't execute in parallel, and when they "jump >>> the queue", further executions are not received. Right? >> >> Right. It's "out of band", not "in parallel". > > OK. Naively I thought that new OOB commands would continue to be > processed "out of band", but it's fine if that capacity is finite. The out-of-band feature is not meant to compete with jobs. It's a narrow solution to a specific problem: robust postcopy migration recovery needs to get in certain commands reliably and quickly, even previously sent commands hog the monitor. A second use has since come up: a 'yank' command to recover from hung network connections by shutting them down. >>> Conceptually, then: There's one input queue and two pending execution >>> slots. One slot is for OOB commands, and the other slot is for >>> traditional commands. >>> >>> The input queue is FIFO, and when we pull a request from the queue, it >>> occupies one of the conceptual slots (normal/oob). If the slot for >>> this command type is already occupied, we block until it's free. >>> >>> Is that the correct mental model? (Even if it's not really even >>> vaguely correct, details-wise.) >> >> Almost. >> >> Consider >> >> - Send execute A >> - Send execute B >> - Send exec-oob C >> - Send execute D >> > > [...] > >> The I/O thread receives in-band A, and puts it into the empty in-band >> queue. >> >> The I/O thread receives in-band B, and puts it into the non-full in-band >> queue. >> >> The I/O thread receives out-of-band C, and exectutes it right away. >> > > Ah, I see where I went wrong. OK, it's crystal clear now. > >> The I/O thread receives in-band C, and puts it into the non-full in-band >> queue. >> >> Meanwhile, the main thread takes in-band A out of the queue, and starts >> executing it. A takes its own sweet time. >> >> Only if the client has more in-band commands in flight than the in-band >> queue can hold, the I/O thread encounters a full queue and suspends >> reading. >> >> qmp-spec.txt: >> >> If the client sends in-band commands faster than the server can >> execute them, the server will stop reading the requests from the QMP >> channel until the request queue length is reduced to an acceptable >> range. >> >> The queue length should probably be specified here. It's 8. >> > > Oh! That's very helpful to know, actually. If you write a patch to amend > the language of the doc, I will give you my R-B. Added to my queue. >> The server limits the in-band request queue length for the same >> robustness reasons it limits JSON token length, token count and nesting >> depth: to not let a malfunctioning client make it consume unlimited >> amounts of memory. Hole (kind of): it doesn't limit output queue >> length. >> > > Understood! > >>> Is there a reason to not always use \xFF ? >> >> It's invalid UTF-8. >> >> "Older versions" should probably read "ancient versions" by now. I >> don't remember offhand which commit fixed the JSON parser to do the >> right thing in all states. >> > > I suppose right now I don't have any code that tries to unjam the > parser, so I suppose it doesn't matter. > > If it did, I guess I would just try to assume that I was allowed to use > the newer method and too-bad to folks trying to talk to ancient stuff. Right. >>>>>> Questions? >>>>> A few :) >>>> More? >>> None of tremendous import, I think, by now. Maybe the GQA stuff, but >>> that can be later. >> Still more? > > Where does matter come from? If energy cannot be created or destroyed, > how did it come to be? "D'où venons-nous? Que sommes-nous? Où allons-nous?" ;)
