We have released version 2.2.1 of the Snowplow Stream Collector, with a fix for a bug identified in 2.1.2 and improvements to the surge protection capability for AWS introduced in 2.0.0.
Fixed issues with anonymous tracking mode
In 2.1.2 we introduced a change that would remove the network_userid from an event if that event was sent by a tracker with the anonymous tracking option enabled. That was to ensure that this identifier would not be inadvertently captured due to the presence of historical cookies on the client machine. However, it caused issues downstream from the collector, resulting in events going to the bad (or failed_events) stream.
This has now been changed to instead use a zero-UUID (00000000-0000-0000-0000-000000000000).
We are also going to be releasing a version 2.1.3 of the stream collector with this fix, for users who cannot yet upgrade to the 2.2.x series but want to use the anonymous tracking feature.
You can find more information about the issue and its lineage in this post: Anonymous Tracking issues in Stream Collector versions 2.1.1, 2.1.2 and 2.2.0 - #2.
Improved surge protection
This release brings a number of improvements to the surge protection feature for AWS, most notably:
- the SQS batch request limit is now being respected (as opposed to just the single-message limit which was the case in previous versions);
- in case a target (Kinesis or SQS) is temporarily unavailable or its throughput allowance is being exceeded, any subsequent re-tries of the write requests will regularly rotate between targets to give them the best chance of succeeding eventually.
You can read more about surge protection in the 2.0.0 release post.