Pipe changelog
Pipe changelog
addpipe.com

Fixed issue with the recording not stopping at the maximum recording time (mrt setting) if the recorder was removed from DOM

 

Fix

 

 

Until now, if the Pipe desktop recording client was removed from the DOM before hitting the maximum recording limit, the recording simply continued forever until the user left or closed the webpage on which the Pipe recording client was embedded on.

The issue has been fixed today when we rolled out new, updated client side (pipe.js) files.

This only occurred when the Pipe recording client was removed using custom JS code. Removing the Pipe recorder with the built in remove function worked as expected and is the recommended way to remove Pipe from a webpage.

Aligned and slightly higher file size limit for uploads

 

Improvement

 

 

Whether you're trying to:

  • upload existing recordings through our desktop recorder
  • upload recordings through the mobile native recording client
  • POST them directly to our private POST API

we have a new slightly raised size limit for the files.

The limit is now 1GiB which is 1024 * 1024 * 1024 = 1.073.741.824 bytes.

Previously, the limit was 1000 MiB which is 1000 * 1024 * 1024 bytes.

The new higher limit means any file shown as 1GB on Linux and macOS (1.000.000.000 bytes) and any file shown as 1GB on Windows (1.073.741.824 bytes) will be accepted by our servers.

Fixed issue with incorrect public URLs sent with the video_copied_s3 webhook for S3 buckets with dots in their name

 

Fix

 

 

If your S3 bucket had a dot in the name, the URLs (url, rawRecordingUrl, snaptshotUrl, filmstripUrl) sent through the video_copied_s3 webhook would be incorrect as they missed the bucket name. This happened across storage services, regardless if you used S3 buckets from AWS or S3 API compatible storage.

Longer but correct time to recover recordings

 

Fix

 

 

Our mechanism that recovers recordings if the user is completely disconnected has been updated to recover the recordings after 15 minutes.

This has been done because our 30 reconnections attempts from the client side can, in theory, take just a bit longer than 12 minutes.

The documentation has been updated with more details.

The rollout of this new mechanism has been completed today across all our regions.

Webhooks timeout reduced to 30 seconds

 

Fix

 

 

Today we've reduced the webhook timeout from 127 seconds to 30 seconds to prevent our virtual workers from being tied up in waiting for responses.

This change should not affect any integration dependent on webhooks as 30 seconds is more than enough to open a TCP connection and POST the tiny amount of data. Also, the platform is not dependent on the response, we just save the status, headers and body to the db, show them in the webhook logs section and delete them a month later.

The extension .3g2 is now fully supported for mobile and desktop uploads

 

Improvement

 

 

We've added the .3g2 extension to our list of extensions that are allowed when uploading from a mobile device using the mobile native recorder or when uploading directly from the desktop.

Here is a list of all the supported extensions (bolded is new):

Desktop Uploads Mobile Uploads
mp4 mp4
mov mov
webm webm
3gp 3gp
3gpp 3gpp
3g2 3g2
avi avi
m4v m4v
ogv ogv
mod mod
qt qt
wmv wmv
mpeg mpeg
mpg mpg
aac aac
m4a m4a
mp3 mp3
wav wav
ogg ogg
wma wma
flac flac
amr amr

Round 2 of improvements to our recording servers

 

Improvement

 

 

This round of changes focuses on one thing in particular: the way we add tasks for our transcoding servers.

Until now, the recording servers relied on the database to pass on the task needed to be done towards the transcoding servers.

With this update, we've implemented direct communication between the recording servers and the transcoding servers.

What does this mean for you?

  • Faster processing (up to 150 ms faster )
  • No more rare random delays (up to 2 minutes) in processing the recordings.

Stay tuned for the next round of updates for our recording servers!

Fixed issue with connecting from behind an ipv6 proxy

 

Fix

 

 

Until today, some visitors using IPv6 might have had issues loading the Pipe Recorder. Those days are gone!

In situations where the user was connecting through IPv6 to a proxy and then to our client delivery server or CDN the X-Forwarded-For header contained IPv6 value which led to a 500 HTTP error instead of a proper JSON response. Parsing the header value (which lead to the error) only happened when the environment's region was set to Auto or Auto-US. From our investigation, this was a rare event. During the past month, it has affected under 100 requests against our precheck mechanism.

This also happened when the X-Forwarded-For header contained malformed information in which case we now just default to the eu region (when the region option is set to Auto on the environment) or us1 (Auto-US).

This is yet another improvement in a series of infrastructure improvements that we're making to the Pipe Platform.

The recording ID from the webhooks payload is now a consistent data type

 

Fix

 

 

The recording ID within a webhook payload was correctly sent as an Integer data type when the recording was processed in real-time (first attempt), but incorrectly sent as a String data type when the webhook was sent as a result of a successfully retried recording (one that failed on its first attempt).

This has been fixed so that the recording ID within any of the possible webhook payloads will always be an Integer data type across the board as it should.

Round 1 of improvements to our recording servers

 

Improvement

 

 

We just deployed part 1 of an ongoing huge update that our recording servers will receive. Here are some of the most important changes we made in the first round:

  • The recording servers can now make multiple connections to our database at the same time.
  • Reduced the number of queries made to our database by implementing a caching mechanism.
  • Code optimization and restructuring.
  • Made changes to lower the number of duplicate recordings even further. This was already a rare occurrence.
  • Better logs with new data to allow us to better keep track of potential issues in the future.

All of these changes mean that our servers will be able to take on much higher recording loads than before for less hardware usage.

Stay tuned for the next round!