--resources="cpus(prod):8; mem(prod):16384; disk(prod):65536" //roles;
4.1.3 Customizing Mesos slave resources and attributes ??
hardcode < physical available ?
Types: scala, range, set
--attributes="datacenter:pdx1; rack:1-1; os:rhel7; pythons:python2,python3" //for scheduling?
containerizer: mesos/docker
Linux资源隔离:
cgroups/cpu and cgroups/mem
filesystem/shared
namespace/pid
posix/disk posix/cpu posix/mem
network/port_mapping
--isolation="cgroups/cpu,cgroups/mem,network/port_mapping" --resources="ports:[31000-32000];ephemeral_ports:[32768-57344]" --ephemeral_ports_per_container=1024 //就是普通的端口映射?
--egress_rate_limit_per_container=12500KB //单位为什么不写成KBps呢?
缺陷容忍
checkpointing
slave recovery
For slave recovery to work, the mesos-slave daemon must reconnect to the master within 75 seconds.
但可以用systemd、Upstart来弥补?
syslog --> logstash/splunk
每个task的log:基于work_dir=/var/lib/mesos
层次:slave -> framework -> executor -> task -> run
frameworks界面:sandbox(临时目录?)
$ mesos execute --command="echo 'Hello, Mesos'" --name=HelloMesos ➥ --master=$(mesos resolve zk://mesos.example.com:2181/mesos)
5.2.3 Using Mesosphere’s mesos-cli tool
/usr/local/bin/mesos
监控Mesos:Nagios?
6.2 Modifying the Mesos master quorum
Framework rate limiting
首先监控,
Service discovery and routing
HAProxy
Mesos-DNS
MesosSchedulerDriver:可以用Python写?
executor
TaskStatus消息
ExecutorInfo... 这不能保证知道如何调用吧?
Mesos完全没有讲到外部是如何访问内部的服务,端口映射?还是说这里的task不对外发布接口?
4.1.3 Customizing Mesos slave resources and attributes ??
hardcode < physical available ?
Types: scala, range, set
--attributes="datacenter:pdx1; rack:1-1; os:rhel7; pythons:python2,python3" //for scheduling?
containerizer: mesos/docker
Linux资源隔离:
cgroups/cpu and cgroups/mem
filesystem/shared
namespace/pid
posix/disk posix/cpu posix/mem
network/port_mapping
--isolation="cgroups/cpu,cgroups/mem,network/port_mapping" --resources="ports:[31000-32000];ephemeral_ports:[32768-57344]" --ephemeral_ports_per_container=1024 //就是普通的端口映射?
--egress_rate_limit_per_container=12500KB //单位为什么不写成KBps呢?
缺陷容忍
checkpointing
slave recovery
For slave recovery to work, the mesos-slave daemon must reconnect to the master within 75 seconds.
但可以用systemd、Upstart来弥补?
syslog --> logstash/splunk
每个task的log:基于work_dir=/var/lib/mesos
层次:slave -> framework -> executor -> task -> run
frameworks界面:sandbox(临时目录?)
$ mesos execute --command="echo 'Hello, Mesos'" --name=HelloMesos ➥ --master=$(mesos resolve zk://mesos.example.com:2181/mesos)
5.2.3 Using Mesosphere’s mesos-cli tool
/usr/local/bin/mesos
监控Mesos:Nagios?
6.2 Modifying the Mesos master quorum
Framework rate limiting
首先监控,
Service discovery and routing
HAProxy
Mesos-DNS
MesosSchedulerDriver:可以用Python写?
executor
TaskStatus消息
ExecutorInfo... 这不能保证知道如何调用吧?
Mesos完全没有讲到外部是如何访问内部的服务,端口映射?还是说这里的task不对外发布接口?