[
https://issues.apache.org/jira/browse/MESOS-4041?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15035618#comment-15035618
]
Guangya Liu commented on MESOS-4041:
------------------------------------
The reason is that frameork does not set the executor id, both mesos-executor
and marathon:
https://github.com/mesosphere/marathon/blob/master/src/main/scala/mesosphere/mesos/TaskBuilder.scala#L162
This will cause optimistic offer phase 1 does not work if no executor id as the
mesos master cannot get executor id to evict.
If we want marathon also work with optimistic offer phase 1, then we may need
to udpate marathon to set executor info even with command executor.
> Default command executor do not have executor_id
> ------------------------------------------------
>
> Key: MESOS-4041
> URL: https://issues.apache.org/jira/browse/MESOS-4041
> Project: Mesos
> Issue Type: Bug
> Reporter: Guangya Liu
>
> I was doing some test with Marathon on top of mesos and found that when using
> mesos command executor, the executor_id is always empty.
> {code}
> "state": "TASK_RUNNING",
> "slave_id": "62c4d3e2-7c80-4d80-a0cd-57b2eced1d81-S0",
> "resources": {
> "ports": "[33505-33505]",
> "mem": 16,
> "disk": 0,
> "cpus": 0.1
> },
> "name": "t1",
> "id": "t1.ac8c4679-98cd-11e5-8b71-823a8274cef3",
> "framework_id": "62c4d3e2-7c80-4d80-a0cd-57b2eced1d81-0001",
> "executor_id": ""
> }
> {code}
> When I test with mesos-executor with command line executor, also no executor
> id.
> {code}
> {
> "unregistered_frameworks": [],
> "frameworks": [
> {
> "webui_url": "",
> "user": "root",
> "used_resources": {
> "mem": 256,
> "disk": 0,
> "cpus": 1
> },
> "unregistered_time": 0,
> "id": "820e082f-be7c-4b59-abc5-9c02f9e8d66d-0000",
> "hostname": "devstack007.cn.ibm.com",
> "failover_timeout": 0,
> "executors": [],
> "completed_tasks": [],
> "checkpoint": false,
> "capabilities": [],
> "active": true,
> "name": "",
> "offered_resources": {
> "mem": 0,
> "disk": 0,
> "cpus": 0
> },
> "offers": [],
> "pid":
> "scheduler-43ee3a42-9302-4815-8a07-500e42367f41@9.111.242.187:60454",
> "registered_time": 1445236263.95058,
> "resources": {
> "mem": 256,
> "disk": 0,
> "cpus": 1
> },
> "role": "*",
> "tasks": [
> {
> "statuses": [
> {
> "timestamp": 1445236266.63443,
> "state": "TASK_RUNNING",
> "container_status": {
> "network_infos": [
> {
> "ip_address": "9.111.242.187"
> }
> ]
> }
> }
> ],
> "state": "TASK_RUNNING",
> "slave_id": "3e0df733-08b3-4883-b3fa-92bdc0c05b2f-S0",
> "resources": {
> "mem": 256,
> "disk": 0,
> "cpus": 1
> },
> "name": "cluster-test",
> "id": "cluster-test",
> "framework_id": "820e082f-be7c-4b59-abc5-9c02f9e8d66d-0000",
> "executor_id": ""
> }
> ]
> }
> ],
> "completed_frameworks": []
> }
> {code}
> This caused the end use can not use http end point to kill some executors as
> mesos require executor id when kill executors.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
