commit 4a99b434851a6020722fe35bc83882a4bcc79169
Author: Greg Althaus <galthaus@austin.rr.com>
Date: Tue Sep 3 09:58:42 2019 -0500
Update to v4.0.20
M go.mod
M go.sum
commit 3ef4f79aebfa65f026bbdb9c79954cd574077508
Author: Victor Lowther <victor.lowther@gmail.com>
Date: Thu Aug 29 15:11:59 2019 -0500
Add token endpoint for machines, and make token TTL handling in the API take things besides a raw number of seconds
M backend/dataTracker.go
M backend/machines.go
M backend/timeParse.go
M clitest/machines_test.go
M clitest/test-data/output/TestUserCli/users.token.rocketskates.scope.all.ttl.cow.action.list.specific.asdgag/stderr.expect
M frontend/machines.go
M frontend/users.go
commit d8fc5da4bf59f35470c05c00d83f1076c7569f8f
Author: Victor Lowther <victor.lowther@gmail.com>
Date: Thu Aug 29 13:18:45 2019 -0500
Move token generation to the Machine model
M backend/machines.go
M backend/renderData.go
commit 13a5ab15d6e42c43ed590d686734e76c55aeaecd
Author: Greg Althaus <galthaus@austin.rr.com>
Date: Thu Aug 29 12:58:41 2019 -0500
Force should be allowed to change stage and bootenv too.
M backend/machines.go
commit fa7d2dca9a3aad99a59a108af36b03fa2c502c4e
Author: Greg Althaus <galthaus@austin.rr.com>
Date: Wed Aug 28 11:07:39 2019 -0500
Add profile-in-profile to the feature-list
M backend/dataTracker.go
M clitest/test-data/output/TestAuth/info.get/stdout.expect
M clitest/test-data/output/TestUserCli/users.token.rocketskates.scope.all.ttl.330.action.list.specific.asdgag/stdout.expect
M clitest/test-data/output/TestUserCli/users.token.rocketskates/stdout.expect
commit 25224995c5bc6108539926518d2daa2968a6d56b
Author: Victor Lowther <victor.lowther@gmail.com>
Date: Tue Aug 27 16:10:16 2019 -0500
Fix a couple of corner cases in log replay
It was possible for log replay to fail due to being unable to
open a file during phase 1 of log replay. This can happen if we are
replaying at the same time as another routine is creating a new WAL
while the system is under heavy load. Since this is an otherwise
harmless condition, do not stop playback when it happens.
Make replayLogs run until it has cleaned out all the logs, and adjust how
we calculate the number of outstanding logs to replay to be tolerant of
potential bad math.
Also make replyLogs die and take the rest of the system if it encounters
a potential dataloss scenario, instead of blithely trying to continue.
Make replay under pressure be more aggressive about pausing everything when
we accumulate too much transaction backlog. Instead of spinning off a
goroutine to wait in that case, stop the whole world right then and wait
for everything to finish processing before resuming operations.
M backend/requestTracker.go
M backend/stack.go
M backend/wal.go
End of Note