GitHub user chjhaijiang opened a pull request:
https://github.com/apache/mesos/pull/184
In CentOS 7, the libevent-devel dependency is needed to be installed for
mesos SSL build.
In the link
(https://github.com/apache/mesos/blob/master/docs/getting-started.md), it
should be a proper instruction once user want to build a SSL mesos. Right now,
it is missed.
You can merge this pull request into a Git repository by running:
$ git pull https://github.com/apache/mesos master
Alternatively you can review and apply these changes as the patch at:
https://github.com/apache/mesos/pull/184.patch
To close this pull request, make a commit to your master/trunk branch
with (at least) the following in the commit message:
This closes #184
----
commit 97bd957ffc45b3484dd82321c5e7fe7f02f9d79c
Author: Neil Conway <[email protected]>
Date: 2016-10-19T23:32:14Z
Changed master to send TASK_UNKNOWN during reconciliation.
Previously, the master would send TASK_LOST in response to explicit
reconciliation requests for (a) unknown tasks at registered slaves and
(b) tasks at unknown slaves (neither registered nor unreachable). The
master will now send TASK_UNKNOWN for these situations if the framework
has the PARTITION_AWARE capability.
Review: https://reviews.apache.org/r/52693/
commit da2ca0f17d9ac9a0248edd18ed2cc7c774e985d4
Author: Neil Conway <[email protected]>
Date: 2016-10-19T23:32:21Z
Renamed a function for clarity.
`Master::reconcile(Framework*, const scheduler::Call::Reconcile&)` and
`Master::reconcile(Slave*, const vector<ExecutorInfo>&, const
vector<Task>& tasks)` are only loosely related. Per discussion on the
development list, using overloading to distinguish these two functions
is confusing. Hence, rename the latter to `reconcileKnownSlave`.
Review: https://reviews.apache.org/r/52719/
commit bf00061b5dd2d5552ad14a206cf8167b70d669c3
Author: Neil Conway <[email protected]>
Date: 2016-10-19T23:32:27Z
Clarified a comment.
Review: https://reviews.apache.org/r/52720/
commit e65b40d48b25ecf45805c8a740a412074da00d1f
Author: Megha Sharma <[email protected]>
Date: 2016-10-20T00:34:19Z
Fixed a bug that causes the fetcher to not chown the sandbox.
Moved the `uri.size() == 0` check in fetcher so that the chown to
task user of stdout/stderr in sandbox directory happens even when
there is no uri to be fetched.
Review: https://reviews.apache.org/r/52828/
commit 09a1cd10278992360c63a77d2712b9d047ce0e67
Author: Megha Sharma <[email protected]>
Date: 2016-10-20T00:37:16Z
Fixed fetcher to not recursively chown the entire sandbox.
Fetcher currently changes the ownership of entire sandbox directory
recursively to the task user and as a result also changes the
ownership of files laid down by other entities in the sandbox, which
leads to unintended side-effects.
Review: https://reviews.apache.org/r/52058/
commit 6880a5f6c82aa5b535ebe2f1b86bad5ca338cb15
Author: Santhosh Kumar Shanmugham <[email protected]>
Date: 2016-10-20T15:54:52Z
Close socket after setting flags on the interface.
Review: https://reviews.apache.org/r/53049/
commit d46807a1c672989b70a33c1fc9355b0bee3855fd
Author: haosdent huang <[email protected]>
Date: 2016-10-20T17:41:14Z
Added non-recursive version of `cgroups::get`.
In most cases, we just want to get the children cgroups instead of
retrieve descendant cgroups recursively. We added an argument to
`cgroups::get` to indicate whether to retrieve cgroups recursively and
made non-recursive retrieve the default behaviour. This patch fixed
some incorrect `TEST_CGROUPS_ROOT` checks as well.
Review: https://reviews.apache.org/r/51031/
commit 4c25f17d87a098a4eb0815156447730cc6111942
Author: haosdent huang <[email protected]>
Date: 2016-10-20T17:44:11Z
Removed the expired TODO about non-recursive version `cgroups::get`.
Review: https://reviews.apache.org/r/51185/
commit fcdc547d720cbb962ec2e498a70fbadd3cf75397
Author: Anindya Sinha <[email protected]>
Date: 2016-10-20T18:05:37Z
Treat failure in destroy of non-existing nested cgroups as a success.
This may happen if certain daemon on the host creates and deletes
nested cgroups inside Mesos containers but the nested cgroups are not
managed by Mesos; we shouldn't fail the destroy. Ultimately the
rationale for this change is that if the goal is to destroy a cgroup
and the end result is that it's gone, then we should consider it a
success even though some nested cgroups are cleaned up out-of-band
by other entities.
Review: https://reviews.apache.org/r/53031/
commit 78d4ec406f7bee61eb5097bca91bf143d2f43f82
Author: Joris Van Remoortere <[email protected]>
Date: 2016-10-18T22:33:09Z
Removed extra 'evolve' implementation from 'api_tests.cpp'.
Review: https://reviews.apache.org/r/53013
commit 7831f1fbace2ae868dd7dc80f4ddca459b9ffe19
Author: Joris Van Remoortere <[email protected]>
Date: 2016-10-18T23:18:25Z
Fixed usage of 'evolve' in master http endpoints.
Review: https://reviews.apache.org/r/53012
commit 2373819dc3e3f8b251526db962eecde23de1545b
Author: Joris Van Remoortere <[email protected]>
Date: 2016-10-19T03:54:41Z
Removed unused tests helper macro 'DEFAULT_CONTAINER_ID'.
Review: https://reviews.apache.org/r/53014
commit e9da9b3bc41aa81c25d36901e52ff1e941fa09e6
Author: Joris Van Remoortere <[email protected]>
Date: 2016-10-18T06:15:21Z
Split mesos test helpers into 'internal' and 'v1' namespaces.
Review: https://reviews.apache.org/r/52976
commit 70e20af17a3f77dd042c8ef437e9f4d890cc40b9
Author: Jiang Yan Xu <[email protected]>
Date: 2016-10-20T21:07:28Z
Revert "Added non-recursive version of `cgroups::get`."
This reverts commit d46807a1c672989b70a33c1fc9355b0bee3855fd.
commit e14c9f7e55f3c9851c11d75149be22bc1967516d
Author: Jiang Yan Xu <[email protected]>
Date: 2016-10-20T21:07:34Z
Revert "Removed the expired TODO about non-recursive version
`cgroups::get`."
This reverts commit 4c25f17d87a098a4eb0815156447730cc6111942.
commit f6b0185024e72b927bc75cfcf0a64c0805ee1b89
Author: Benjamin Bannier <[email protected]>
Date: 2016-10-21T16:16:04Z
Ensured agent is recovered in SlaveEndpointTests.
The tests in this case did either assume that e.g., requests for agent
endpoints could always reliably be satisfied by a not yet fully
recovered agent, or did not explicitly observe the recovery and
instead relied on just settling the clock.
Make sure agents are actually fully recovered, even if not strictly
required right now. Also actually observe agent recovery calls like
done elsewhere.
Review: https://reviews.apache.org/r/52818/
commit 7e93d9d17b9f79b35b0e0c5b882a51a6e5aa1db2
Author: Joris Van Remoortere <[email protected]>
Date: 2016-10-20T20:36:28Z
Alphabetized 'evolve' and 'devolve' files.
Review: https://reviews.apache.org/r/53076
commit b75aa8ada2d8ce5ff98fa8d99f04e45527b5953e
Author: Joris Van Remoortere <[email protected]>
Date: 2016-10-20T20:37:00Z
Redefined helper macros with existing 'create' functions.
Review: https://reviews.apache.org/r/53077
commit 89e7d362e0cf57758bfa44328c5ea2917251d189
Author: Johannes Unterstein <[email protected]>
Date: 2016-10-21T19:07:07Z
Updated getting started for Ubuntu 16.04.
To follow the getting started instructions to install Mesos on
Ubuntu 16.04, it is needed to install one more dependency.
This closes #173
commit 9225ac294099afe09d85bf95931b83b78a8739c7
Author: Neil Conway <[email protected]>
Date: 2016-10-21T21:13:02Z
Fixed typo in log message.
Review: https://reviews.apache.org/r/52721/
commit b7d8c29cdcb29a3a0ca1551e73745cd28c796729
Author: Neil Conway <[email protected]>
Date: 2016-10-21T21:13:08Z
Changed master to add `FrameworkInfo` to agent reconcilation.
When an agent re-registers with a master that already knows about that
agent, the master needs to reconcile its view of the state of the agent
with the agent's current state. For any task that the master thinks
should be on the agent but isn't included in the ReregisterSlaveMessage,
the master does a reconcilation with the agent to find the task's
current state.
This commit changes adds the `FrameworkInfo` for any possibly missing
tasks to the master -> agent reconciliation message. This is useful
because the agent can consult the `FrameworkInfo` during reconciliation:
this will shortly be used to make agent reconciliation behave
differently for partition-aware frameworks.
Review: https://reviews.apache.org/r/52722/
commit 0312bc159d6ed375feebb217af7b5ab456e31bc9
Author: Neil Conway <[email protected]>
Date: 2016-10-21T21:13:17Z
Changed agent to send TASK_DROPPED during reconciliation.
If a framework attempts to launch a task but the launch message is
dropped after it reaches the master but before it reaches the slave, the
failed launch will be detected during master <-> agent reconciliation
when the agent re-registers. Previously, the agent would generate a
TASK_LOST status update for such dropped tasks; now it will generate
TASK_DROPPED if the framework is partition-aware.
Note that we'll only send TASK_DROPPED if the agent is running a
sufficiently recent version of Mesos (>= 1.1.0). That means that in a
mixed cluster where the master has been upgraded to Mesos 1.1 but some
of the agents have not been, a partition-aware framework might still see
TASK_LOST in this situation.
Review: https://reviews.apache.org/r/52723/
commit 0645e7d9d5354375f4020ef8c8b30f053c6ccca5
Author: Neil Conway <[email protected]>
Date: 2016-10-21T21:13:24Z
Refactored some code into a separate function.
Review: https://reviews.apache.org/r/52740/
commit c4b69ec1b8ac798f3f68bf116f596a2d09aeaa15
Author: Neil Conway <[email protected]>
Date: 2016-10-21T21:13:30Z
Changed agent to send TASK_DROPPED for task launch failures.
If the agent cannot launch a task due to a variety of possible error
conditions, we now send TASK_DROPPED to partition-aware frameworks
rather than TASK_LOST.
Review: https://reviews.apache.org/r/52746/
commit 976a1ebcd8ecc336a91dcc384a3c5f11df4faec1
Author: Neil Conway <[email protected]>
Date: 2016-10-21T21:13:36Z
Changed description of TASK_GONE.
Review: https://reviews.apache.org/r/52801/
commit b46df16ddee455c6ddb9171d7ffccfa0defbeb54
Author: Neil Conway <[email protected]>
Date: 2016-10-21T21:13:40Z
Added a new slave metric, "tasks_gone".
Review: https://reviews.apache.org/r/52802/
commit 01aa3ba4a043e3aff04a632008f65a6ba33f8dcb
Author: Neil Conway <[email protected]>
Date: 2016-10-21T21:13:45Z
Changed agent to send TASK_GONE.
The agent previously sent TASK_LOST updates for tasks that are killed
for various reasons, such as containerizer errors or QoS preemption. The
agent now sends TASK_GONE to partition-aware frameworks instead.
Review: https://reviews.apache.org/r/52803/
commit 27eb6b7faed68e83a6b713c4f69c958abba78100
Author: Neil Conway <[email protected]>
Date: 2016-10-21T21:18:46Z
Avoided passing `TimeInfo` by value.
Although this is likely to remain small in practice, passing by const
reference should be preferred until there is a reason not to.
Review: https://reviews.apache.org/r/53099/
commit 61a0a10d4a83695a76115385b2401e6059062a8a
Author: Neil Conway <[email protected]>
Date: 2016-10-21T21:18:52Z
Fixed bug when marking agents unreachable after master failover.
If the master fails over and an agent does not re-register within the
`agent_reregister_timeout`, the master marks the agent as unreachable in
the registry and sends `slaveLost` for it. However, we neglected to
update the master's in-memory state for the newly unreachable agent;
this meant that task reconciliation would return incorrect results
(until/unless the next master failover).
Review: https://reviews.apache.org/r/53097/
commit f9dfbdac74588883390fab8dfcce9528708ac83a
Author: Neil Conway <[email protected]>
Date: 2016-10-21T21:18:59Z
Tweaked test expectation.
`WillOnce` is more accurate than `WillRepeatedly`.
Review: https://reviews.apache.org/r/53098/
----
---
If your project is set up for it, you can reply to this email and have your
reply appear on GitHub as well. If your project does not have this feature
enabled and wishes so, or if the feature is enabled but not working, please
contact infrastructure at [email protected] or file a JIRA ticket
with INFRA.
---
