Is there a memory limit to the Task Queue on Google App Engine. I'm specifically concerned with the Go runtime, but it would be nice to get answers on all runtimes if someone can provide them.
The tasks are executed by the same app instance(s) as the regular requests, only tasks are allowed to run longer. So the same memory limits apply (also subject to the task queue specific limits and quota, which might also eat into the instance memory).
You might chose to direct tasks to a dedicated module to which you assign an instance class with more memory (more powerful as well), if memory consumption is a concern.
But since the max instance class memory size is currently 1G I suspect your instance will most likely hit the 'soft private memory limit' and be killed before loading a full 1G file into memory :)
A "task" is essentially represented by a URL that's stored away for later delivery to an instance of your app. The representation of a task is independent of language, unless you use a stringified, language-specific serialization of something as a value.
If by memory limit you mean "how much (where much = count*size) task queue stuff can I have pending?," the answer is spelled out in the Task Queue section of the quotas document.
If you're asking how big a single task can be, that'll depend on the memory size of your instances, since you'll need enough memory to construct a task before enquing it.
For task processing the app instances need enough memory to accept and process a task, or enough memory to accept and process many concurrently, if your app is configured to accept multiple simultaneous requests. How much memory that takes beyond accepting the URL is basically up to how the app is coded.
Related
I have a few requests that needs to use extensive amount of memory i.e. 40 MB more than other requests.
At the default of 10 max concurrent requests using a F1 auto-scaling instance, it can potentially use 400+ MB, which is way more than the 130MB-ish system memory it has available. There is no memory utilization setting in the yaml file, so I wonder what can be done to prevent such situations.
Google App Engine don't have any memory utilization beside Python Garbage Collection.
My Advise is,
Try to release memory as soon as response
Try to optimize memory usage on that part, may be you need to use other service to help solving memory usage problem, eg. file serving via Google Storage, etc...
Scale up instance to F2 which more suitable for production, but you still need to optimize your memory usage for higher usage.
We are using AppDynamics and VisualVM to monitor our application heap memory usage. We see similar graph as stated in these questions - this and this.
the red boxes show idle system heap usage - peaks are seen only when system is in idle state and are even observed when no application is deployed.
the green arrow points to actual application in use state - When system is in use, we see relatively very less heap usage being reported.
Based on the clarifications in other SO questions, if we say it is due to garbage collection, why would GC not occur during application use? When system is idle, we see system objects like java.land.String, byte[], int[] etc. getting reported in AppDynamics, but how to find who is responsible for creating them?
Again, in the heap dumps taken during idle state, we see only 200MB out of 500MB memory used, when the server has dedicated -Xmx4g configuration.
How should we make sense of these observations?
On analyzing the heap dump taken during system idle state, we only see various WebAppClassLoaders holding instances of different library classes.
This pattern is also explained in official blogs of APM experts like Plumbr and Datadog as a sign of healthy JVM where regular GC activity is occurring and they explain that it means none of the objects will stay in memory forever.
From Plumbr blog:
Seeing the following pattern is a confirmation that the JVM at question is definitely not leaking memory.
The reason for the double-sawtooth pattern is that the JVM needs to allocate memory on the heap as new objects are created as a part of the normal program execution. Most of these objects are short-lived and quickly become garbage. These short-lived objects are collected by a collector called “Minor GC” and represent the small drops on the sawteeth.
Scenario-
I am running B* instances on App Engine. I've a background ETL related task(written in python) scheduled as a cron job on App Engine.
When time arrives, cron initiates a http request to start the task and runs without returning a response till the task gets completed.
When task was executing, it was typically consuming "X" MB of RAM. After the task got finished and returned 200 OK, App Engine Instance monitoring is still showing "X" MB of RAM in use.
Please help me understand the following -
If an instance is running only one task and after completing it, when will memory get freed that was consumed by this task?
Do I need to run gc.collect() to call the garbage collector explicitly to free up RAM ?
The only way to free up RAM is to restart the instance ?
PS: This is not at all related to NDB, my task is taking input from Bigquery, performing some ETL operation and then streaming it to Bigquery.
From my observations with an app using lots of memory for StringIO operations:
explicitly calling gc.collect() didn't noticeably help (I even suspected for a while that I actually have memory leaks, but it wasn't the case)
the memory is not freed after each and every request, but, if the instance remains alive long enough without running out of memory it does eventual
appears to be freed now and then. Easy to test - just increase the time between requests to reduce the free memory draining rate. But I couldn't figure out a usable pattern. Note that I observed this only after upgrading to B2 instances, my B1 instances were running out of memory too fast, I never noticed a freeing event with them.
using an instance class with more memory (which I tried as a workaround for my instances eventually running out of memory) helped - the memory appeared to be freed more often. It might be because these instances also have a faster CPU (but that's just guesswork).
There are a few questions on StackOverflow describing similar memory issues for tasks when using ndb on app engine. Here is one example.
The issue is that app engine doesn't clear the ndb context cache upon the conclusion of a task so context cache continues to hog your memory long after the task completes.
The solution is to not use or clear the context cache during your tasks. Here are a few ways:
Bypass caching with key.get(use_cache=False)
Call ndb.get_context().clear_cache() at appropriate times.
Disable caching for all entities of a kind by adding _use_cache = False to your model definition.
I saw this Python question: App Engine Deferred: Tracking Down Memory Leaks
... Similarly, I've run into this dreaded error:
Exceeded soft private memory limit of 128 MB with 128 MB after servicing 384 requests total
...
After handling this request, the process that handled this request was found to be using too much memory and was terminated. This is likely to cause a new process to be used for the next request to your application. If you see this message frequently, you may have a memory leak in your application.
According to that other question, it could be that the "instance class" is too small to run this application, but before increasing it I want to be sure.
After checking through the application I can't see anything obvious as to where a leak might be (for example, unclosed buffers, etc.) ... and so whatever it is it's got to be a very small but perhaps common mistake.
Because this is running on GAE, I can't really profile it locally very easily as far as I know as that's the runtime environment. Might anyone have a suggestion as to how to proceed and ensure that memory is being recycled properly? — I'm sort of new to Go but I've enjoyed working with it so far.
For a starting point, you might be able to try pprof.WriteHeapProfile. It'll write to any Writer, including an http.ResponseWriter, so you can write a view that checks for some auth and gives you a heap profile. An annoying thing about that is that it's really tracking allocations, not what remains allocated after GC. So in a sense it's telling you what's RAM-hungry, but doesn't target leaks specifically.
The standard expvar package can expose some JSON including memstats, which tells you about GCs and the number allocs and frees of particular sizes of allocation (example). If there's a leak you could use allocs-frees to get a sense of whether it's large allocs or small that are growing over time, but that's not very fine-grained.
Finally, there's a function to dump the current state of the heap, but I'm not sure it works in GAE and it seems to be kind of rarely used.
Note that, to keep GC work down, Go processes grow to be about twice as large as their actual live data as part of normal steady-state operation. (The exact % it grows before GC depends on runtime.GOGC, which people sometimes increase to save collector work in exchange for using more memory.) A (very old) thread suggests App Engine processes regulate GC like any other, though they could have tweaked it since 2011. Anyhow, if you're allocating slowly (good for you!) you should expect slow process growth; it's just that usage should drop back down again after each collection cycle.
A possible approach to check if your app has indeed a memory leak is to upgrade temporarily the instance class and check the memory usage pattern (in the developer console on the Instances page select the Memory Usage view for the respective module version).
If the pattern eventually levels out and the instance no longer restarts then indeed your instance class was too low. Done :)
If the usage pattern keeps growing (with a rate proportional with the app's activity) then indeed you have a memory leak. During this exercise you might be able to also narrow the search area - if you manage to correlate the graph growth areas with certain activities of the app.
Even if there is a leak, using a higher instance class should increase the time between the instance restarts, maybe even making them tolerable (comparable with the automatic shutdown of dynamically managed instances, for example). Which would allow putting the memory leak investigation on the back burner and focusing on more pressing matters, if that's of interest to you. One could look at such restarts as an instance refresh/self-cleaning "feature" :)
My instances all now start at 140m and average just under 200. If left long enough they start hitting 240m. However my question is more about the memory being used right after a fresh instance is booted up. I store nothing on the instances. Every request fetches stuff from memcache and datastore and I don't use singletons.
All I have are classes and a lot of static resources that deploy with the instances. I use JSPs extensively (if that makes a difference).
Thanks for any assistance!
I'm going from memory here, since I have used Java on App Engine for a few years. This may be stale.
The JVM doesn't like to release memory. If an instance gets created, and services a request, the memory watermark goes up. Garbage collection may 'free' part of that memory up in the sense of it being available for reuse, but the high watermark on process memory doesn't necessarily go down. A subsequent request may need allocations that aren't available as free chunks, so the watermark on memory goes up again. If the app isn't configured to serve multiple request simultaneously, memory use follows something like a sigmoid curve. If multiple requests are being processed simultaneously, the watermark is raised further.
That said, a common cause of unexpected memory growth is queries that retrieve more rows than are necessary, with filtering happening in the app.
But without more information, your specific case is impossible to diagnose.
I believe I figured out why my project was taking up an ever increasing amount of ram. I happen to have a lot of static resources in my project and it would appear that these static resources all get loaded into the frontend instance memory (probably for speed). I managed to free up huge amounts of memory by moving my static resources off of my primary application servers.