Heroku has great utilities to manage web dynos. When we do a small release we use Preboot to gracefully switch over web dynos over to the new release.
However, if you want to scale down your web dynos, there is no graceful way to do so. Ideally we can mark a web dyno so that Heroku Scheduler would stop giving it new requests for 3 minutes before sending SIGTERM.
So if i have 6 dynos running, but outside of office hours I want to ps:scale web=2, it sends SIGTERM to the other 4. If there was a straggler in slow times routed to one of those 4 (instead of the remaining 2), they'd experience a crash resulting from SIGTERM.
Is there a way to gracefully scale down web dynos?
Correctly handling SIGTERM is the recommended approach for graceful shutdown, see https://devcenter.heroku.com/articles/dynos#graceful-shutdown-with-sigterm. Please note, this the same shutdown process that occurs during preboot too. So when your process receives a SIGTERM, it should stop responding to requests, and finish responding to its current requests, after 30 seconds, if the process hasn't exited, then the process is forcefully SIGKILL'd. There is no builtin way to perform rolling scaledown, you'd have to write your own interface on the Heroku API to achieve this.
Related
I'm familiar with Heroku's policy on dyno restarts (once every 24 hours, if underlying hardware fails, etc).
I've also looked elsewhere on Stackoverflow and found questions like
Controlling Heroku's random dyno restarts.
Our apps are good from a web perspective – sessions are handled via external database and multiple dynos balance the load perfectly. Restarts aren't an issue there. The issue is workers. For example, our worker receives a message and begins processing a job. It's 99% done, awaiting some final asynchronous request to return, and it receives SIGTERM. Before it can even clean up the job, the process is killed. The code can handle local cleanup for a job that needs to be restarted, but external services can't really be part of the transaction.
For example, if a report was built and an email was sent, but some 3rd async operation didn't complete before SIGTERM, I can't really roll back that transaction. For hardware failures or other rare events, it's understandable that a multi-step transaction could get truncated, but with Heroku's policy it seems that I need to assume this will happen at least once a day. Can anyone help me get a better grip on this problem?
I'm testing an app with a worker and a web dyno on Heroku free tier and I'd like to keep the worker alive to be able to execute background tasks while letting the web dyno idle. By default they both go idle in 30 mins even if I have things queued on the worker.
I understand there're ways to keep the web dyno alive (and with that the worker as well), and there're ways to keep the web alive while scaling down the worker. However I'd need the worker alive and the web in idle.
I tried running a recurring job on the worker which would
Restart the dyno.
Scale the dyno down and then back up.
Both approaches worked (as in they restarted and scaled the dyno correctly) but the worker dyno would still idle after 30 mins (as if it's dependent on the web dyno). Edit: yep, that's pretty much the case as explained here: https://devcenter.heroku.com/articles/free-dyno-hours#dyno-sleeping
I could do this form the outside but it seems I'd have to constantly check for the state since a new restart doesn't seem to give me 30 mins headway. I'd also have to expose the API key which I'd like to avoid.
If I've gotten you right, you're trying to stop the web dyno and leave the worker dyno alive.
You could do that by going to the Resources tab:
And then in the 'web' section:
Press the pencil, toggle it off and press 'Confirm'.
As a workaround I currently remove the web dyno and explicitly enable it when I need it. As explained here:
Worker-only Free dynos do not sleep since they do not respond to web
requests.
My workaround was to just create two apps that deploy automatically from the same repository. Then, all you would need to do is enable the worker dyno for one and the web dyno for the other.
I am new to Go and I am using go routines in my app in Heroku, which are long (up to 7 minutes), and cannot be interrupted.
I saw that the auto scaler sometimes kills the Heroku dyno which is running the routine. I need a way of running this routine independently from the dynos so I know that it will not get shutdown. I read articles and still don't understand how to perform a go routine in a background worker. It is hard for me to believe I am the only one experiencing this.
My go routines use my redis database.
Could someone please point me to an example of how to setup a background worker in heroku for go and how to send my go routine to that worker?
Thank you very much
I need a way of running this routine independently from the dynos so I
know that it will not get shutdown.
If you don't want to run your worker code on a dyno then you'll need to use a different provider from Heroku, like Amazon AWS, Digital Ocean, Linode etc.
Having said that, you should design your workers, especially those that are mission critical, to be able to recover from a shutdown. Either to be able to continue where they left off or to start over. Heroku's dyno manager restarts the dynos at least once a day but I wouldn't be surprised if the other cloud providers also restart their virtual instances once in a while, probably not once a day but still... And even if you decide to deploy your workers on a physical machine that you control and never turn off, you cannot prevent things like hardware failure or power outage from happening.
If your workers need to perform some task till it's done you need to make them be aware of possible shutdowns and have them handle such scenarios gracefully. Do not ever rely on a machine, physical or virtual, to keep running while your worker is doing it's job.
For example if you're on Heroku, use a worker dyno and make your worker listen for the SIGTERM signal, after your worker receives such a signal...
The application processes have 30 seconds to shut down cleanly
(ideally, they will do so more quickly than that). During this time
they should stop accepting new requests or jobs and attempt to finish
their current requests, or put jobs back on the queue for other worker
processes to handle. If any processes remain after that time period,
the dyno manager will terminate them forcefully with SIGKILL.
... continue reading here.
But keep in mind, as I mentioned earlier, if there is an outage and Heroku goes down, which is something that happens from time to time, your worker won't even have those 30 seconds to clean up.
I have two web dynos in my heroku app, and at times get a dyno automatic restart (as per heroku policy). Is the function that was going on during the restart automatically restored in the new restarted dyno? If not, is there a way I can control this restart?
Is the function that was going on during the restart automatically restored in the new restarted dyno?
no
If not, is there a way I can control this restart?
no
What you can do, is trap the SIGTERM signal that is sent to your process 10 seconds before it is SIGKILLed. This would give you time to finish current computation, stop taking web requests, do cleanup, etc. More details on the process is in the Heroku Devcenter.
Very similar question to Is it feasible to run multiple processeses on a Heroku dyno?, or Running Heroku background tasks with only 1 web dyno and 0 worker dynos except I'm talking about a Ruby on Rails app.
Context:
I understand that it's encouraged to separate worker and web dynos... but I'm still testing and don't want to pay the expense. Especially because with my app, all the web requests pretty much happen either in the AM or in the PM, and during the whole middle of the day (and also middle of the night), literally nothing is happening.
I'd like the web dyno to run two types of background processing on the "downtime":
A recurring, long-running task every day (mailings)
An asynchronous, long-running task that is triggered when a user performs a certain action (it's a mailer)
I've done quite a bit of reading on this, but this is my first time doing anything asynchronous, so I wanted to ask the community a couple of questions just to ensure what I'm trying to do is feasible.
Questions
How do I do activity #1 for free?
To put it bluntly... considering my context above, if I use Heroku's Scheduler add-on, this runs a one-off dyno which I'll be charged for since I use NewRelic now to constantly ping my web dyno so it never actually sleeps meaning my one web dyno is my free dyno. Is there another way of doing this with the one web dyno that, in the middle of the night, won't be processing any requests? Alternatively, is there a way to tell New Relic to ping except at certain times, which will also then allow me to spin up a one-off dyno but still be within my free dyno hours?
For activity #2, I'm thinking of using Delayed Jobs, but how do I tell Delayed Jobs to delay until end of user 1's session, and then run mailer for user 1, but then pause again if user 2 sends a request, and then when user 2 is finished, start where left off on user 1's mailer, and then do user 2's mailer... and so forth? I think the root of the challenge here is that from what I've read, Delayed Jobs needs to be started with a script. But I'm not going to be at my computer starting a script all the time. How do I make the start (and the queue as illustrated in the question) something that happens automatically?
Would love even just point me directional pointers on what methods/ what considerations, etc.
I'm going to check out a nifty gem https://github.com/brandonhilkert/sucker_punch to do this. According to the author, it was written specifically to use Heroku's single dyno for hobby websites that have no need to spin up another dyno. It basically creates another thread.
FYI also, there is an add-on link that allows sucker_punch to do recurring tasks, called https://github.com/facto/fist_of_fury