I have an app with Carrierwave on Heroku. On a page, I have 2 forms: 1 ajax form for uploading a picture and 1 normal form for additional information needed to create the object. Suppose my Carrierwave mount is :picture, every time the ajax form is submitted, the picture is saved temporarily into the public folder and its path is returned as :picture_cache. The second form then uses that to know which picture to be created with the new object on the second request. This works fine for a single dyno.
Different dynos don't know about each other's filesystems. Thus if the request to submit the 2nd form doesn't hit the same dyno as the request of the first form, it can't find the image.
Has anyone tackled this problem?
i use a custom model and store all files, including tmp ones, in mongodb. the uploads are marked as tmp. ones the models is 'saved' i simply remove the 'tmp' flag. in this way all nodes see all images all the time. it's pretty crazy that the carrierwave default is to cache in ./tmp since many multi-node configuration would see this issue (unless the balancer implements session affinity).
here is my model and controller, etc: https://gist.github.com/3161569
you have to do some custom work in the form:
save every file posted, no matter what
relay the posted file id in a hidden field
on save look for a file and/or previously uploaded id
make the model associations
this approach, although it isn't 'magic' also gives the following awesome side effects:
you have one process running jobs in the background to thumbnail the images vs. spinning up image_magick whenever a user hits 'submit' (which is a serious DOS vector, esp on memory limited hosts like heroku)
you can migrate images to s3 in the background, hourly, whatever, and the uploads simply have a new url (in this case the controller need to issue a permanant redirect if it notices this). this is really nice because you can keep 'em in the db for dev, staging, etc. and migrate some, or all, uploads onto s3 whenever without changing any upload or view code.
Related
I've just discovered that Heroku doesn't have long-term file storage so I need to move to using S3 or similar. A lot of new bits and pieces to get my head around so have I understood how direct upload to S3 using CarrierWave-direct and then processing by delayed_job should work with my Rails app?
What I think should happen if I code this correctly is the following:
I sign up to an S3 account, set-up my bucket(s) and get the authentication details etc that I will need to program in (suitably hidden from my users)
I make sure that direct upload white lists don't stop cross-domain from preventing my uploads (and later downloads)
I use CarrierWave & CarrierWave-direct (or similar) to create my uploads to avoid loading up my app during uploads
S3 will create random access ('filename') information so I don't need to worry about multiple users uploading files with the same name and the files getting overwritten; if I care about the original names I can use metadata to store them.
CarrierWave-direct redirects the users browser to an 'upload completed' URL after the upload from where I can either create the delayed_job or popup the 'sorry, it went wrong' notification.
At this point the user knows that the job will be attempted and they move on to other stuff.
My delayed_job task accesses the file using the S3 APIs and can delete the input file when completed.
delayed_job completes and notifies the user in the usual way e.g. an e-mail.
Is that it or am I missing something? Thanks.
You have a good understanding of the process you need. To throw one more layer of complexity at you---you should wrap all of it in rails new(er) ActiveJob. ActiveJob simply facilities background processing inside rails via the processor of your choosing (in your case DelayedJobs). Then, you can create Jobs via a rails generator:
bin/rails g job process_this_thing
Active Jobs offers a few "rails way" of handling jobs...but, it also allows you to switch processors with less hassle.
So, you create a carrierwave uploader (see carrierwave docs). Then, attach that uploader to a model. For carrierwave_direct you need to disassociate the file field from your models form and move the file field to its own form (use the form url method provided by carrierwave-direct).
You can choose to upload the file, then save the record. Or, save the record and then process the file. The set-up process is significantly different depending on which you choose.
Carrierwave and carrierwave-direct know where to save the file based on the fog credentials you put in the carrierwave initializer and by using the store_dir path, if set, in the uploader.
Carrierwave provides the uploader, which define versions, etc. Carrierwave_direct facilities uploading direct to your S3 bucket and processing versions in the background. Active Jobs, via DelayedJobs, provides the background processing. Fog is the link between carrierwave and your S3 bucket.
You should add a boolean flag to your model that is set to true when carrierwave_direct uploads your image and then set to false when the job finishing processing the versions. That way, instead of a broken link (while the job is running and not yet complete) your view will show something like 'this thing is still processing...'.
RailsCast is the perfect resource for completing this task. Check this out: https://www.youtube.com/watch?v=5MJ55_bu_jM
In my application, I have a textarea input where users can type a note.
When they click Save, there is an AJAX call to Web Api that saves the note to the database.
I would like for users to be able to attach multiple files to this note (Gmail style) before saving the Note. It would be nice if the upload could start as soon as attached, before saving the note.
What is the best strategy for this?
P.S. I can't use jQuery fineuploader plugin or anything like that because I need to give the files unique names on the server before uploading them to Azure.
Is what I'm trying to do possible, or do I have to make the whole 'Note' a normal form post instead of an API call?
Thanks!
This approach is file-based, but you can apply the same logic to Azure Blob Storage containers if you wish.
What I normally do is give the user a unique GUID when they GET the AddNote page. I create a folder called:
C:\TemporaryUploads\UNIQUE-USER-GUID\
Then any files the user uploads at this stage get assigned to this folder:
C:\TemporaryUploads\UNIQUE-USER-GUID\file1.txt
C:\TemporaryUploads\UNIQUE-USER-GUID\file2.txt
C:\TemporaryUploads\UNIQUE-USER-GUID\file3.txt
When the user does a POST and I have confirmed that all validation has passed, I simply copy the files to the completed folder, with the newly generated note ID:
C:\NodeUploads\Note-100001\file1.txt
Then delete the C:\TemporaryUploads\UNIQUE-USER-GUID folder
Cleaning Up
Now. That's all well and good for users who actually go ahead and save a note, but what about the ones who uploaded a file and closed the browser? There are two options at this stage:
Have a background service clean up these files on a scheduled basis. Daily, weekly, etc. This should be a job for Azure's Web Jobs
Clean up the old files via the web app each time a new note is saved. Not a great approach as you're doing File IO when there are potentially no files to delete
Building on RGraham's answer, here's another approach you could take:
Create a blob container for storing note attachments. Let's call it note-attachments.
When the user comes to the screen of creating a note, assign a GUID to the note.
When user uploads the file, you just prefix the file name with this note id. So if a user uploads a file say file1.txt, it gets saved into blob storage as note-attachments/{note id}/file1.txt.
Depending on your requirement, once you save the note, you may move this blob to another blob container or keep it here only. Since the blob has note id in its name, searching for attachments for a note is easy.
For uploading files, I would recommend doing it directly from the browser to blob storage making use of AJAX, CORS and Shared Access Signature. This way you will avoid data going through your servers. You may find these blog posts useful:
Revisiting Windows Azure Shared Access Signature
Windows Azure Storage and Cross-Origin Resource Sharing (CORS) – Lets Have Some Fun
I am working on a Rails web application, running on a Heroku stack, that handles looking after some documents that are attached to a Rails database object. i.e. suppose we have an object called product_i of class/table Product/products, and product_i_prospectus.pdf is the associated product prospectus, where each product has a single prospectus.
Since I am working on Heroku, and thus do not have root access, I plan to use Amazon S3 to store the static resource associated with product_i. So far, so good.
Now suppose that product_i_attributes.txt is also a file I want to upload, and indeed I want to actually fill out information in the product_i object (i.e. the row in the table corresponding to product_i), based on information in the file product_i_attributes.txt.
In a sentence: I want to create, or alter, database objects, based on the content of static text files uploaded to my S3 bucket.
I don't actually have to be able to access them once they are in the bucket strictly speaking, I just need to create some stuff out of a text file.
I have done something similar with csv files. I would not try to process the file directly at upload as it can be resource intensive.
My solution was to upload the file to s3 and then call a background job method(delayed_job, resque, etc.) that processed the csv after upload. You could then call a delete after the job processed to remove the file from s3 if you no longer needed it after processing.
For Heroku this will require that you add a worker (if you don't already have one) to process the background jobs that will process the text files.
Take a look at the aws-sdk-for-ruby gem. This will allow you to access your S3 bucket.
Im desiging an app which allows users to upload images (max 500k per image, roughly 20 images) from their hard drive to the site so as to be able to make some custom boardgames (e.g. snakes and ladders) in pdf formate. These will be created with prawn instantly and then made available for instant download.
Neither the images uploaded nor the pdfs created need to be saved on my apps side permanently. The moment the user downloads the pdf they are no longer needed.
Heroku doesn't support saving files to the system (it does allow to the tmp directory but says you shouldnt rely on it striking it out for me). I'm wondering what tools / services I should be looking into to get round this. Ive looked into paperclip, I'm wondering if this is right for this type of job.
Paperclip is on the right track, but the key insight is you need to use the S3 storage backend (Paperclip uses the FS by default which as you've noticed is no good on Heroku). It's pretty handy; instead of flushing writes out to the file system, it uses the AWS::S3 gem to upload them to S3. You can read more about it in the rdoc here: http://github.com/thoughtbot/paperclip/blob/master/lib/paperclip/storage/s3.rb
Here's how the flow would work:
I'd let your users upload their multiple source images. Here's an article on allowing multiple attachments to one model with paperclip: http://www.cordinc.com/blog/2009/04/multiple-attachments-with-vali.html.
Then when you're ready to generate the PDF (probably in a background job, right?), what you do is download all the source images to somewhere in tmp/ (make sure the directory is based on your model id or something so if two people do this at once, the files don't get stepped on). Once you've got all the images downloaded, you can generate your PDF. I know this is using the file system, but as long as you do all your filesystem interactions in one request or job cycle, it will work, your files will still be there. I use this method in a couple production web apps. You can't count on tmp/ being there between requests, but within one it's reliably there.
Storing your generated PDF on S3 with paperclip makes sense too, since then you can just hand your users the S3 URL. If you want you can make something to clear the files off every so often if you don't want to pay the S3 costs, but they should be trivial.
Paperclip sounds like an ideal candidate. It will save images in RAILS_ROOT/public/system/, which is both persistent and private (shouldn't be able to be enumerated on shared hosting).
You can configure it to produce thumbnails of your images if you wish.
And it can remove the images it manages when the associated model is destroyed - after your user downloads their PDF, and you delete the record from the database.
Prawn might not be appropriate, depending on the complexity of the PDFs you need to generate. If you have $$$, go for PrinceXML and the princely gem. I've had some success with wkhtmltopdf, which generates PDFs from a Webkit render of HTML/CSS - but it doesn't support any of the advanced page manipulation stuff that Prince does.
I decided to use Amazon S3 for document storage for an app I am creating. One issue I run into is while I need to upload the files to S3, I need to create a document object in my app so my users can perform CRUD actions.
One solution is to allow for a double upload. A user uploads a document to the server my Rails app lives on. I validate and create the object, then pass it on to S3. One issue with this is progress indicators become more complicated. Using most out-of-the-box plugins would show the client that file has finished uploading because it is on my server, but then there would be a decent delay when the file was going from my server to S3. This also introduces unnecessary bandwidth (at least it does not seem necessary)
The other solution I am thinking about is to upload the file directly to S3 with one AJAX request, and when that is successful, make a second AJAX request to store the object in my database. One issue here is that I would have to validate the file after it is uploaded which means I have to run some clean up code in S3 if the validation fails.
Both seem equally messy.
Does anyone have something more elegant working that they would not mind sharing? I would imagine this is a common situation with "cloud storage" being quite popular today. Maybe I am looking at this wrong.
Unless there's a particular reason not to use paperclip I'd highly recommend it. Used in conjunction with delayed job and delayed paperclip the user uploads the file to your server filesystem where you perform whatever validation you need. A delayed job then processes and stores it on s3. Really, really easy to set up and a better user experience.