$ git log --no-merges --oneline b19d028de0a6e9ca281afeb76cea2544b9edd839..f420c4b9e1535170fc229db97ee8ac32374020b1
87d3730 orchestrator: Add test debugging
dccbce3 It's possible for a new session to start before the previous session has completely finished cleaning up in the fake dispatcher.  Therefore, in the cleanup, don't set the current open session to nil without checking if the current open session has changed.  This should prevent some false negative test assertions where it seems like a session has never been established even though it has.
ff551b0 agent: Resolve deadlock in run
8edbb92 Adding PlatformFilter for scheduling
8c72dcf trigger tick when resources could be freed up
5f8f5ac scheduler should re-evaluate pending tasks after any task removal
78f240e [protos] Adding Platforms field to TaskSpec.Placement
fde7e71 Fix bug where we could not cross-sign an RSA certificate with an ECDSA certificate and vice versa.
51309e8 Connections to a local dispatcher can’t really be closed, so a session can’t really be restarted because closing a session just closes the connection.  When this happens, it just starts up another session without closing the previous one.
c21a196 Add details field to log message proto
6b36068 Adding test case to verify vip allocation in ingress mode
54c7200 Fixing leak in Ingress VIP allocation on port publish/remove
f9f376f Update `identity/doc.go` to be coherent with the code
5d77af7 Temporarily vendor fork of etcd
dac5c30 ca: Delay between retries when retrieving certs
1474fd1 node: Third proposed fix for cert renewal logic race
444b318 Adding test case for ingress network
019ec5d Fixing regression from #2142
423c661 Add previously invalid targets to TestSecretValidation
e11952a update tests
d14fe90 allow secret references to be full paths
745e1ea Avoid writing to Network.Spec in allocatePools
55bf1ae node: Use filepath.Join properly
9363459 agent: avoid race condition after failing task startup
4e307ca Fix broken link in README
89edef8 Minor doc changes.
ec2e6c5 allocator: Make updatePortsInHostPublishMode check for host publish mode, not ingress
0818c04 allocator: Standardize on label/continue syntax when comparing VIPs with networks
8f96e51 allocator: Factor specNetworks assignment into a function
891b020 allocator: Rename PortsAllocatedInHostPublishMode
cad1bb5 store: Simplify Batch
f8ef88b Adding test case for service update
b7672d9 Fixing service update
333fd8f When polling that the key is decrypted after it is encrypted, wait until the node is a leader first.  Otherwise, it may poll to early, get the raft store before it has caught up with all the logs (including the one that changes the CA key to an encrypted version) and hence think that the key has been decrypted when it hasn't, yet.
b652e9b docs: change readme badge to svg
5af584f Only log messages about updating root CA in CA server if it's the not the first time a cluster object has been set.
c605cee Actually enable CheckQuorum
1481b02 raft: Return a more useful error message when quorum is lost
69e1189 node: Support older CAs which don't set Node.Role
3a485cc validate health check parameters and refactor some code
ee44570 Bump the timeout for the agent test TestSessionRestartedOnNodeDescriptionChange in order to avoid spurious test failures.  Also include a little more info on why the test failed, if it fails.
a6e08a0 Watch API monitors local store for change
7a427bf rename storeapi to watchapi
fb8dc43 allocator: Remove redundant nil check
1c666c1 raft: Don't log errors removing members based on snapshot
213dca6 In the dispatcher, start watching for cluster changes before sending the first session message, because it's possible to miss a cluster update just between when the message is being sent and when we start watching for cluster updates.