Robert Marianski
Robert Marianski
When enqueueing tiles, the list of tiles should be sorted highest zoom levels first when they are grouped. Because we'll process them in the order of the message, this will...
The rawr tile work has been done using the boto3 library, and the sqs queue was updated to use boto3 in [this pr](https://github.com/tilezen/tilequeue/pull/290). The rest of the code should be...
Everywhere else we have timing information in milliseconds. The tilequeue timing was the original and just remained in seconds, but this should be updated to be in milliseconds to be...
The `table-sources` rawr config should be lifted out of the `source` config. The table sources themselves aren't inherent in how to fetch the data, and it will make the `source`...
Errors during coord ack appeared in the latest tilequeue run. These two types appear releated. ``` Traceback (most recent call last): File \"/usr/local/lib/python2.7/dist-packages/tilequeue/worker.py\", line 524, in __call__ queue_msg_handle, done =...
Audit the codebase to ensure that threads use separate boto sessions.
It might be possible to introduce a race condition to produce stale tiles. 1. a tile exists in the toi, and is generated on s3 2. a diff affects that...
Sqs drops messages after some time. When this occurs, nothing removes the corresponding tile from the in-flight set. This is problematic because once we get into this state we would...
At the moment, most of the formatters expect wkb. We should either extend them or modify them to take in a shapely object directly. All of them load the geometry...