Made the agent verify resource compatibility in checkpointResources(). In handling CheckpointResourcesMessage, the agent should first make sure the checkpointed resources are compatible before it syncs local disk state.
Review: https://reviews.apache.org/r/51866 Project: http://git-wip-us.apache.org/repos/asf/mesos/repo Commit: http://git-wip-us.apache.org/repos/asf/mesos/commit/b24d99b6 Tree: http://git-wip-us.apache.org/repos/asf/mesos/tree/b24d99b6 Diff: http://git-wip-us.apache.org/repos/asf/mesos/diff/b24d99b6 Branch: refs/heads/master Commit: b24d99b68bae1445a5948b68c0130d7c1c08ca71 Parents: e65933c Author: Jiang Yan Xu <xuj...@apple.com> Authored: Tue Sep 13 14:53:34 2016 -0700 Committer: Jiang Yan Xu <xuj...@apple.com> Committed: Mon Sep 19 14:59:11 2016 -0700 ---------------------------------------------------------------------- src/slave/slave.cpp | 20 ++++++++++++++++++-- 1 file changed, 18 insertions(+), 2 deletions(-) ---------------------------------------------------------------------- http://git-wip-us.apache.org/repos/asf/mesos/blob/b24d99b6/src/slave/slave.cpp ---------------------------------------------------------------------- diff --git a/src/slave/slave.cpp b/src/slave/slave.cpp index 7f99e46..4ced3e7 100644 --- a/src/slave/slave.cpp +++ b/src/slave/slave.cpp @@ -2736,6 +2736,19 @@ void Slave::checkpointResources(const vector<Resource>& _checkpointedResources) return; } + // This is a sanity check to verify that the new checkpointed + // resources are compatible with the agent resources specified + // through the '--resources' command line flag. The resources + // should be guaranteed compatible by the master. + Try<Resources> totalResources = applyCheckpointedResources( + info.resources(), + newCheckpointedResources); + + CHECK_SOME(totalResources) + << "Failed to apply checkpointed resources " + << newCheckpointedResources << " to agent's resources " + << info.resources(); + // Store the target checkpoint resources. We commit the checkpoint // only after all operations are successful. If any of the operations // fail, the agent exits and the update to checkpointed resources @@ -5051,8 +5064,11 @@ Future<Nothing> Slave::recover(const Result<state::State>& state) } // This is to verify that the checkpointed resources are - // compatible with the slave resources specified through the - // '--resources' command line flag. + // compatible with the agent resources specified through the + // '--resources' command line flag. The compatibility has been + // verified by the old agent but the flag may have changed during + // agent restart in an incompatible way and the operator may need + // to either fix the flag or the checkpointed resources. Try<Resources> totalResources = applyCheckpointedResources( info.resources(), checkpointedResources);
