Archived from the Transloadit support forum. Originally posted on October 18, 2011; details may be outdated, see our docs for current behavior. Original thread
This morning we started getting this error when uploading videos.
CONNECTION_ERROR: There was a problem connecting to the upload server
JSONP request status: timeout
Nothing has changed on our end and it was working fine yesterday.
Hey there Chris,
we are currently experiencing very high load. It will work again within 5 minutes.
Sorry for the inconvenience. : /
Kind regards,
Tim
Hi Tim,
We are currently in closed beta, however soon we will be launching live and our load alone could be very considerable.
I need an honest answer to an honest question which is will your company be able to scale to handle the additional load without us experiencing any downtime?
Best,
Chris
Hey Chris,
we are running fine normally - it’s just that one user imports tens of thousands of images at the same time (!) sometimes. We have temporarily disabled his account and got in touch with him.
However, to give you an honest answer, please let me know how much load you anticipate per day.
Kind regards,
Tim
PS: Things have been back to normal for the past minutes already.. This only caused a downtime of 4 minutes.
Hey Tim,
On our side, I first heard about the problem at 9:48am EST, just for your guys information. So it was definitely more then a 4 minute down time for us.
We have a twitter/facebook like app (I hate saying that, but it’s a the best general descriptionI can give) so uploading video is an option, but I wouldn’t see us uploading 10,000 videos ever at once. However in the next few months if things blow up as we plan on then we could be seeing multiple large videos being uploaded at once all through out the day.
But video is by no means our “main” feature it is just one of many features for when users post.
Best,
Chris
Hey Chris,
sounds cool!
Our problem is that we currently do not have queues implemented. That means that just the machine with the lowest load (we also use some other metrics) is chosen for the job. If many jobs get spawned at once, then the machine with the lowest load receives all of them, which could cause it to become swapping or unresponsive altogether.
However, in the next two months we’ll work hard on our infrastructure to make it scale better (implement queues, separate uploading from encoding, among other things) and implement some cool new features like resumable uploads as well.
We would be pleased to keep you on board to grow together.
If you need a discount on your bill please let me know.
Kind regards,
Tim
Hey Chris,
the infrastructure upgrades have been deployed a couple weeks ago.
I am closing this ticket for now. Please reopen it as you feel the need to.