adding connect and read timeouts for firehose client

[ryandeivert]

to: @jacknagz or @chunyong-lin
cc: @airbnb/streamalert-maintainers
size: small
resolves: #478

Background

The vast majority of our lambda execution time is spent trying to send data to firehose via boto3. This can cause timeouts, etc. The AWS kinesis team suggested shortening the connect and read timeouts to cause retries to happen faster. It's our understanding that if the connect or read takes anything over a few seconds, it's likely not going to succeed at all. Therefore, it's advantageous to just have the attempt killed at retried sooner.

Changes

• As suggested, this changes the connect and read timeouts for boto3 via settings in a botocore.client.Config.
• Adding logger statement that will track firehose FailedPutCount values that are greater than 0.

[coveralls]

Coverage increased (+0.004%) to 97.611% when pulling 0b92417 on ryandeivert-shorten-fh-timeout into af596ea on master.

[jacknagz]

Thank you for this!