I have a request that takes more than 30 seconds and it breaks.
What is the solution for this? I am not sure if I add more dynos this will work.
Thanks
You should probably see the Heroku devcenter article regarding this, as the information will be more helpful, here's a small summary:
To answer the timeout question:
Cedar supports long-polling and streaming responses. Your app has an initial 30 second window to respond with a single byte back to the client. After each byte sent (either recieved from the client or sent by your application) you reset a rolling 55 second window. If no data is sent during the 55 second window your connection will be terminated.
(That is, if you had Cedar instead of Aspen or Bamboo you could send a byte every thirty seconds or so just to trick the system. It might work.)
To answer your dynos question:
Additional concurrency is of no help whatsoever if you are encountering request timeouts. You can crank your dynos to the maximum and you'll still get a request timeout, since it is a single request that is failing to serve in the correct amount of time. Extra dynos increase your concurrency, not the speed of your requests.
(That is, don't bother adding more dynos.)
On request timeouts:
Check your code for infinite loops, if you're doing something big:
If so, you should move this heavy lifting into a background job which can run asynchronously from your web request. See Queueing for details.
Related
I'm doing website optimisations using Google's Pagespeed Insights to test improvements. Among the high-priority fix suggestions, is this:
Reduce server response time
In our test, your server responded in 2.1 seconds.
I read the 'helpful' doc linked in this section, and now I'm really confused.
Is the server response time the DNS response, the time to first-byte, or a combination? Is it purely a server-side thing, or could this be affected by, for example, a slow JavaScript resource or ready events in the DOM?
My first guess would have been that it's the time taken from the moment the request was issued, to the 1st byte received from the server, however Google's definition is not quite that:
(from this page https://developers.google.com/speed/docs/insights/Server)
Server response time measures how long it takes to load the necessary
HTML to begin rendering the page from your server, subtracting out the
network latency between Google and your server. There may be variance
from one run to the next, but the differences should not be too large.
In fact, highly variable server response time may indicate an
underlying performance issue.
To take 2.1 seconds would suggest to me that your application/webserver is buffering it's output, so all your server side processing is happening before it sends the content. If you don't buffer then the html can begin being sent to the browser more quickly which may help, however you lose the ability to do things like change response headers late in your logic.
First of all I'm developing my own C# library for controlling Philips Hue, which means I'm not using the official SDK. (I'm guessing that the SDK will make sure you won't have any problems)
I'm a little confused about the limitation in the Core concepts page in the API, which states:
We can’t send commands to the lights too fast. If you stick to around 10 commands per second to the /lights resource as maximum you should be fine. For /groups commands you should keep to a maximum of 1 per second.
I intend to respect this limitation, but does the limitation still apply when you are performing GET requests on the /lights resource, or is it only for sending actual commands with PUT requests to /lights/<id>/state that change the state of the light? Same question goes for the /groups resource.
Also is it even possible to damage anything by sending too many requests, or will it just take longer to get all responses?
Edit:
My overall question is: How should I understand the API limitation?
A more specific sub-question is: Should I wait 100 ms before sending another /lights command, relative to when I received a response, or relative to when I sent the previous command?
Another sub-question is: Should I consider this limitation only when using PUT requests on e.g. /lights/<id>/state, or on all request types GET/PUT/POST/DELETE
I don't know if anything was changed in firmware updates, but I have discovered that the bridge might not be so simple as you would think, and that the API description isn't very clear.
I've done a little testing while running firmware 01009914.
The bridge seems to have some kind of queue of incoming commands. I sent {"bri":254} to a group 9 times and 1 final command of {"bri":1}. From the first command to when the light is actually dimmed, takes roughly 3-4 seconds. Each time I sent a command the bridge replied almost instantly with success token.
I did the same small tests sending other commands, 10 of each JSON object:
{"bri":254} 3-4 seconds
{"on":true, "bri":254} 6-7 seconds
{"on":true, "bri":254, "alert":"none", "effect":"none"} 12-13 seconds
This actually shows that each change of attributes takes roughly 0.3 seconds for the bridge to handle.
I will claim that for each attribute we change, the bridge takes about 300 ms to finish, and the limitation of commands should be understood as: As long as you stick with changing one attribute of a group each second, you should be fine.
Note: I only tried with one group consisting of three lights, and I don't know if the bridge actually does have a queue of incoming commands, and in case it does have a queue, I don't know what the limit of items in it is.
Edit:
Now we have some official clarification of the Hue System Performance.
I'm fairly certain that the 10 commands per second is a guideline to prevent failure of the Bridge, and is a technical limitation of the hardware. Any more than that and you're apt to overload the bridge. I believe this applies to commands as well as requests.
Both approaches are reasonable. For laziness' sake, you could wait for 100ms to send a response, but I would only rely on this method if you don't plan on any other interactions with the Bridge.
I consider this limitation on all request types.
You won't damage anything if you send commands too fast. However, if you send commands too fast the bridge might become unresponsive and/or some messages can be ignored.
When it comes to the bridge, the way I think of it is that the bridge is more or less single threaded, so it works best if you make sure you don't send the next command before the previous one has returned.
In practice we've found that this works much better than waiting a fixed time between each request. In fact, you can pretty much send commands as fast as you want as long as you wait for the previous one to finish.
When you send a command to the bridge, the bridge has to then send it to the lamps through Zigbee. Since it's a mesh network in some cases the message has to make a couple of hops from lamp to lamp before it reaches the target. Depending on how many lamps you have and how many hops the signal needs to take, this can take a while. Also, it's possible that some messages randomly take much longer than others.
In general the system is not designed to handle very fast changes, but if you keep the above in mind you can make many cool effects :)
My app has an API that users can request data. Sometimes that data takes time to process and is breaking my code.
I need a solution for this and I was thinking in using delayed_job but I'm not sure how this works. If the user makes a request, I need to give him an answer. Even if I process the data in background, the call still needs to wait until the job returns.
What is the solution for this? I am not sure how to do it.
Thanks
Heroku has a 30 second timeout, which is why your requests are failing (Probably H12 or H13 in your heroku logs).
There are three methods to work around this.
Keep the connection open by sending blank data.
You'll need to respond within the first 30 seconds and every 55 seconds after that. Use the time in between to process the data. Sending spaces should not affect the ability of the browser to read the response.
Callback
Have the user provide a callback URL in the initial request. When you finish processing the data, hit the callback url with your response.
Polling
As suggested by Codeglot, you can provide the user with a key. To check on their request, they can ping your server with that key.
Tell the user that their data is being processed and will be available shortly. Youtube, Vimeo, Facebook, Twitter, they all do this.
I am using Heroku to host a small app. It's running a screenscraper using Mechanize and Nokogiri for every search request, and this takes about 3 seconds to complete.
Does the screenscraper block anyone else who wants to access the app at that moment? In other words, is it in holding mode for only the current user or for everyone?
If you have only one heroku dyno, then yes it is the case that other users would have to wait in line.
Background jobs are for cases like yours where there is some heavy processing to be done. The process running rails doesn't do the hard work up-front, instead it triggers a background job to do it and responds quickly, freeing itself up to respond to other requests.
The data processed by the background job is viewed later - perhaps in a few requests time, or whenever the job is done, loaded in by javascript.
Definitely, because a dyno is single threaded if it's busy scrapping then other requests will be queued until the dyno is free or ultimately timed out if they hang for more than 30 seconds.
Anything that relies on an external service would be best run through a worker via DJ - even sending an email, so your controller puts the message into the queue and returns the user to a screen and then the mail is picked up by DJ and delivered so the user doesn't have to wait for the mailing process to complete.
Install the NewRelic gem to see what your queue length is doing
John.
If you're interested in a full service worker system we're looking for beta testers of our Heroku appstore app SimpleWorker.com. It's integrated tightly into Heroku and offers a number of advantages over DJ.
It's as simple as a few line of code to send your work off to our elastic worker cloud and you can take advantage of massive parallel processing because we do not limit the number of workers.
Shoot me a message if you're interested.
Chad
We have a memory intensive processing for certain functionality and we would like to limit the number of parallel requests to this processing. We are able to configure by using "Work Managers" in WebLogic and putting a limit on the number of threads for that servlet.
For example, if we put maximim thread limit as 3, then if there are 10 parallel requests; 7 requests are in queue. There could be situations where these the requests waiting in queue could take up to 30-40 minutes to be processed. We did simple testing and the received page cannot be displayed due to timeout after 15 mins and received the message after 1 hour.
Does any one know if there is a setting in WebLogic to increase/decrease timeout and avoid page cannot be displayed?
Appreciate if any one has any thoughts around this.
Does any one know if there is a setting in WebLogic to increase/decrease timeout and avoid page cannot be displayed?
There might be something but I actually didn't check as it would be a bad advice anyway. By looking for this, you are trying to solve the wrong problem here. A browser is just not made for long-running process like the one you are describing (>30mn) even if you don't mind the user waiting (not mentioning that he could refresh the page and queue more and more jobs).
So, the right answer here is in my opinion: use asynchronism, this is the perfect use case. When the user clicks on the button, send a JMS message to a queue (or create a Quartz job) and send the user a page with a request ID telling him to come back later. When the processing is done, update the status somewhere and make the status/result available to the user. Really, the user experience will be better doing this and you'll face less problems than with a browser.
1) Use some other tool (not browser) like WGET where you can control timeout parameter (--timeout).
2) Why do you use HTTP? Use message driven beans and send message JMS to that and don't care about time outs.
Perhaps quartz can do what you need? Start a job and check in on it as you need to?