Throttle Not Working

I’ve recently installed this on my UNRAID server using linuxserver/duplicati docker.

Duplicati - 2.0.5.1_beta_2020-01-18 is the version it says.

I’m trying to setup a max upload of 10Mbps but no matter what I set, it keeps maxing out my upload connection. Even setting it to 100Kbs.

Any ideas? I’ve tried also enabling download throttle (since there is an old issue and that was recommended) but didn’t change anything.

Welcome to the forum @Fiala06

You could try setting below Option to 1. That helps with some destination types. What’s yours?

  --asynchronous-concurrent-upload-limit (Integer): The number of concurrent
    uploads allowed
    When performing asynchronous uploads, the maximum number of concurrent
    uploads allowed. Set to zero to disable the limit.
    * default value: 4

EDIT: Found the option you listed. I’ve set it to 1 and still maxing out my upload.

Thank you. I’m using OneDrive for the destination. I’m not sure where to define that option but I have noticed its the same for Google Drive.

Working fine here on 2.0.5.1 to Google Drive. How are you measuring? Here are Duplicati and Task Manager. Actual network traffic is a bit bursty, but OneDrive v2 looks bumpier without config tweaks.

image

EDIT: Note my setting is 100 KBps, which is 800 Kbps (bytes versus bits) – did you account for that?

The GUI explains the adjustment better:

image

Directly from my router, Unraid server and individual docker monitoring. All point to the same thing, duplicati.

Then let’s see what Duplicati sees, to study its perspective. This is long, but at least do some of this.
Clearly I can’t observe your system, but it seems like you would be able to look at Duplicati, correct?

What does Duplicati’s status bar say (similar to my screenshot), is this at asynchronous-concurrent-upload-limit of 1, and which destination are you using? What happens if you change the throttle rate?

If you would rather not impact your regular backup, you can set up a test backup to do settings tests, sometime when the regular backup is not running. Note that you can still throw off the schedule, e.g. extremely low throttles may make the test job run over, which might impact the regular job start time.

Since you seem router-savvy, I’d note that smoother flows could be created by using router facilities, because that’s a router specialty. Duplicati should be able to do average rates well though. For QoS:

came at the end of a technical dive proposing a OneDrive-specific tuning option to reduce burstiness.
Ultimately the OneDrive setting sounds like it wasn’t used because the router method was doing well.

In addition to viewing the Duplicati status bar (like I showed) to see what it thinks its uploading rate is, detailed results information on whole file uploads is available in a couple of other ways. No-setup is at About → Show log → Live → Information and watch the files go up. Most interesting are large dblocks whose size is configurable in Options (Remote volume size). Watch times, see if rate is too high after setting up the other suggested options, possibly at lower values. You now asked for 2 MBytes, got 3.4, however settings questions remain. Regardless, what happens if you ask for 1 MByte? A slower rate?

Another way to see the upload information to compute a rough rate is in <job> → Show log → Remote where the dblock rate should be the main factor because the dindex files for dblocks are much smaller:

This is kind of rough and requires you to do the math, but a better reading requires looking in log file, e.g.

2021-02-13 21:23:24 -05 - [Information-Duplicati.Library.Main.BasicResults-BackendEvent]: Backend event: Put - Started: duplicati-b7c6d668bcc8b414a887d64cda47db334.dblock.zip.aes (4.92 MB)
2021-02-13 21:24:15 -05 - [Profiling-Duplicati.Library.Main.Operation.Backup.BackendUploader-UploadSpeed]: Uploaded 4.92 MB in 00:00:50.9390276, 98.81 KB/s
2021-02-13 21:24:15 -05 - [Information-Duplicati.Library.Main.BasicResults-BackendEvent]: Backend event: Put - Completed: duplicati-b7c6d668bcc8b414a887d64cda47db334.dblock.zip.aes (4.92 MB)

I was running a 5 MB remote volume size to make sure I got a good flow of dblocks with small total data.
Here, Information level time has seconds, not minutes as in live log. Doing my own math, I can compute seconds as 51, and estimate bytes as 5 * 1024 * 1024 = 5242880. 102802/second nears 100 KB target.

Options would be log-file=<path> and log-fle-log-level=information. Retry is similar but shows retry action. Profiling is huge output. You can instead use log-file-log-filter to filter interesting lines, e.g. *UploadSpeed:

2021-02-14 10:27:42 -05 - [Profiling-Duplicati.Library.Main.Operation.Backup.BackendUploader-UploadSpeed]: Uploaded 49.90 MB in 00:08:32.0315811, 99.80 KB/s
2021-02-14 10:27:43 -05 - [Profiling-Duplicati.Library.Main.Operation.Backup.BackendUploader-UploadSpeed]: Uploaded 43.06 KB in 00:00:00.8976540, 47.97 KB/s
2021-02-14 10:28:07 -05 - [Profiling-Duplicati.Library.Main.Operation.Backup.BackendUploader-UploadSpeed]: Uploaded 2.32 MB in 00:00:24.2611805, 97.93 KB/s
2021-02-14 10:28:08 -05 - [Profiling-Duplicati.Library.Main.Operation.Backup.BackendUploader-UploadSpeed]: Uploaded 22.31 KB in 00:00:01.0062725, 22.17 KB/s
2021-02-14 10:28:09 -05 - [Profiling-Duplicati.Library.Main.Operation.Backup.BackendUploader-UploadSpeed]: Uploaded 6.45 KB in 00:00:00.9460525, 6.82 KB/s

The default-size dblock got throttled down, otherwise it would have been about 700 KBytes/s on my line.
Mine is Windows 10, Google Drive, GUI throttle at 100 KBytes, and 1 concurrent upload to keep it simple.
If you’re seeing different results, let’s figure out where things went different, starting from Duplicati’s data.

I did backups to WebDav before and there the throttling worked as expected. I recently had to switch and moved to OnedDrive. The issue of throttling not working seems to only impact OneDrive. I now configured OneDive as follows:

–fragment-size=983040
–fragment-retry-count=10
–asynchronous-concurrent-upload-limit=1

It seems that after each fragment uploaded the speed is checked. So with the 983k fragements the upload only reaches about 1MB/s and then that fragement is finished and it will then check configured upload limit and either wait or continue.

There is another issue with OneDrive and not using the Retry-After header: OneDrive respect "retry-after" header · Issue #4438 · duplicati/duplicati · GitHub

I’m not sure whether you’re trying to make another real-time upload happy or just avoid hogging. Hogging can probably be avoided by throttling down, but the minimum burstiness probably comes from Microsoft:

Upload bytes to the upload session

the size of each byte range MUST be a multiple of 320 KiB (327,680 bytes).

Do you think that’s relevant? Issues exist (as you can see from GitHub). Volunteers are much needed.

My main goals was to not mess up the 4 MS Teams/Zoom sessions we have with the current lockdown. So I wanted minimum impact on the upload of my cable. So I was playing around with that.

My 983040 fragment size is 3x the 320 KB (327680 bytes) so that should be fine. This causes my peakes to be within my limits.

I will check if I can contribute to the retry-after issue.

That’s good to hear, and you have a bit more room to play with if you ever require smaller peaks.
I’m not terribly expert in all the code internals, I’m afraid (GitHub has people, but not enough), but

is what your option dials back to a throttling-friendlier peak. I don’t know what other backends do.
I thought possibly that could be investigated – someday in the future when hotter issues are few.

Thank you. Pull requests are appreciated. Duplicati now has a fixed retry-delay between its tries.
Use exponential backoff for retries #2399 requests what some destinations tell their apps to use.

Any pointers to where in the code the retries are managed?

Generally my approach is to start from the option, then look for a run-together version of it. This leads to:

however Microsoft Graph backends seem to have an additional level that I can’t explain, but will point to:

There is a defined interface (which gets expanded on for backends with additional capabilities) defined at

https://github.com/duplicati/duplicati/blob/master/Duplicati/Library/Interface/IBackend.cs

and I’m not sure how much unwanted isolation that might add going between the code areas. If this gets deeper, you might want to ask in Developer category of the forum, which some developers might visit…

EDIT:

I’m not sure if it fits or will help here, but my hints about the lower-level HTTP response handling are here.

I was the one who originally wrote most of the OneDrive v2 (and other Microsoft Graph backends), so I can provide a bit of context here.

The Microsoft graph backend specific retry logic comes from the OneDrive concept of ‘upload sessions’. Large files need to be uploaded to OneDrive in smaller pieces, so if a file is too large, a single IBackend.Put() call will cause these backends to create an upload session and then perform multiple calls to upload pieces of that file. By default, these fragments are ~10 megabytes. If any of those individual requests fail, these backends handle their own retry internally rather than attempting to preserve the state of the upload session between multiple Put() calls from the Duplicati infrastructure.

I wonder if this ‘upload session’ behavior is also part of why throttling isn’t working as expected (and if this is why network usage is somewhat ‘bursty’). I believe the way Duplicati handles automatic throttling is by introducing a Stream wrapper which performs throttling by only allowing bytes to be read at the throttled rate (see ThrottledStream). However, because the upload session reads the full fragment size into a buffer and then uploads it directly, the throttling will apply only to the process of reading and filling the buffer and not to the actual upload. One way to work around this might be to set a small --fragment-size so that even while the individual fragments are sent unthrottled, they are small enough to be spaced out more to match the desired throttling state. (Though would only work if Duplicati applies the throttling to reading the source files on disk for uploading - if that isn’t how it applies throttling this might not really do anything. I think that is what the following code is doing in BackendUploader though:

// A download throttle speed is not given to the ThrottledStream as we are only uploading data here
using (var fs = File.OpenRead(item.LocalFilename))
using (var ts = new ThrottledStream(fs, m_initialUploadThrottleSpeed, 0))
using (var pgs = new ProgressReportingStream(ts, pg => HandleProgress(ts, pg, item.RemoteFilename)))
    await streamingBackend.PutAsync(item.RemoteFilename, pgs, cancelToken).ConfigureAwait(false);

)

Code fixes to improve throttling behavior might include:

  • Having the Microsoft graph backends automatically scale the fragment size to perform that workaround automatically (this is probably the easiest but least desirable fix).
  • Adding a method on ThrottledStream (which these backends can try to call explicitly if the stream they get is a ThrottledStream) which will read as many bytes as possible without pausing and then return, rather than continuing to read and buffer the entire fragment size. (It may still need to pause in some cases, but hopefully would pause at most once per buffer returned.)
  • The best option is probably to change the logic used by the individual upload session fragments to write the fragments using a ThrottledStream (e.g., using a StreamContent with a ThrottledStream over the fragmentBuffer instead of directly using a ByteArrayContent on line 464 and similarly using a ThrottledStream on line 547 to write to the request stream). It might be possible to write some sort of custom Stream implementation that wraps the input stream and does the fragmenting / limiting directly (e.g., instead of reading into an intermediate byte buffer), but that would be a bit more complex - in particular, I don’t know if Duplicati manages throttling from a central place (e.g., there is a global counter of bytes written / read) or whether those limits are maintained independently on each ThrottledStream.

The backoff / retry strategy comes from the OneDrive documentation here. Specifically:

  • Use an exponential back off strategy if any 5xx server errors are returned when resuming or retrying upload requests.
  • For other errors, you should not use an exponential back off strategy but limit the number of retry attempts made.

When I initially wrote this, I didn’t realize that OneDrive sometimes adds Retry-After headers. (It looks like they are documented here though.) A OneDrive specific fix for this would be to look for this header in MicrosoftGraphBackend.CheckResponse() and just insert a Thread.Sleep() call to wait that amount of time before continuing. A better fix might be to push that logic up into the OAuthHttpClient / OAuthHelper classes (so any backend using them automatically gets RetryAfter support, though I don’t know if that would have any other side effects).

1 Like

Actually, thinking more about the substream approach, I think it would actually be pretty straightforward. I’ll try to take a look at it.

1 Like

I don’t know the high level throttling design, but Fix bandwidth throttling inconsistency comments:

In code the throttling generally works backwards than what one would expect.
The streams that are throttled are the streams that operate on files from the local machine.

That matches what I’ve seen in the code (the code I pasted in above shows the ThrottledStream wrapping the file on disk which is passed as the input to PutAsync()), in which case I hope that for the HttpClient version of the code the change I’m making to pass a wrapper over the original stream helps. However, on Mono, these backends use the same OAuthHelper and AsyncHttpRequest classes as most other backends instead of the newer HttpClient, and in that implementation the request stream needs to be written to, which I suspect means it gets throttled while filling some in memory buffer rather than while sending bytes over the wire.

I’ve opened the following PR to try to address this:

However, whether this actually impacts throttling might depend on the internals of some .NET HTTP wrappers. (I think the HttpClient code path is more likely to help, which is the default only on non-Mono environments, so on Linux this might not affect things much.)

Forcing a smaller fragment size might still be the best option.

1 Like