[MESOS-9210] Mesos v1 scheduler library does not properly handle SUBSCRIBE retries

• Type: Bug
• Status: Resolved
• Priority: Major
• Resolution: Done
• Affects Version/s: 1.5.1, 1.6.1, 1.7.0
• Fix Version/s: 1.5.3, 1.6.2, 1.7.2, 1.8.0
• Labels: integration mesosphere
• Target Version/s: 1.5.3, 1.6.2, 1.7.2
• Sprint: Mesosphere Sprint 2018-29, Mesosphere RI-6 Sprint 2018-30, Mesosphere RI-6 Sprint 2018-31, Mesos Integration RI10 Sp 38
• Story Points: 3
• Assignee: Till Toenshoff
• Reporter: Vinod Kone
• Shepherd: Alex R
• Created: 05/Sep/18 21:17
• Updated: 29/Apr/19 09:26
• Resolved: 30/Jan/19 15:07

Description

After the authentication related refactor done as part of https://reviews.apache.org/r/62594/, the state of the scheduler is checked in `send` (https://github.com/apache/mesos/blob/master/src/scheduler/scheduler.cpp#L234)  but it is changed in `_send` (https://github.com/apache/mesos/blob/master/src/scheduler/scheduler.cpp#L234). As a result, we can have 2 SUBSCRIBE calls in flight at the same time on the same connection! This is not good and not spec compliant of a HTTP client that is expecting a streaming response.

We need to fix the library to either drop the retried SUBSCRIBE call if one is in progress (as it was before the refactor) or close the old connection and start a new connection to send the retried SUBSCRIBE call.