我正试图通过魔法在本地主机上安装OpenStack和NovaLXD,它一直在超时.(在家进行开发/学习)
我的硬件很古老..。这是一个旧的戴尔poweredge 2850 (大约2004年),我从易趣。它有双单核Xeons,启用HT (4逻辑),16 It和6xU 320驱动器。PERC控制器是油炸的,但在原始SCSI模式下工作很好(我用软件raid (LVM +MD)设置)。它“应该”足够了,即使最终的结果是有点迟缓,考虑到魔术推荐16 up和2核。虽然他们也推荐SSD,但我在几次尝试中没有看到过多的硬盘活动。
经过几次尝试微调一些事情,我被困在这里,我认为,只要‘一点点’的时间,将使这一切顺利完成。连续2次的尝试导致了同样的事情,看起来一切看起来都会成功地完成,除了这种不耐烦的魔术/柔术。
我的召唤日志的尾部是:
2018-12-10 19:09:48,717 [INFO] conjure-up/openstack-novalxd - common.py:47 - Waiting for deployment to settle.
2018-12-10 20:45:05,352 [ERROR] conjure-up/openstack-novalxd - juju.py:745 - DEBUG:root:cinder/0 workload status is maintenance since 2018-12-11 02:45:40Z
DEBUG:root:glance/0 workload status is maintenance since 2018-12-11 02:47:53Z
DEBUG:root:keystone/0 workload status is maintenance since 2018-12-11 02:56:09Z
DEBUG:root:neutron-api/0 workload status is maintenance since 2018-12-11 02:42:12Z
DEBUG:root:neutron-gateway/0 workload status is maintenance since 2018-12-11 02:42:46Z
DEBUG:root:nova-cloud-controller/0 workload status is maintenance since 2018-12-11 03:01:57Z
DEBUG:root:nova-compute/0 workload status is maintenance since 2018-12-11 02:59:49Z
DEBUG:root:openstack-dashboard/0 workload status is maintenance since 2018-12-11 03:07:32Z
DEBUG:root:rabbitmq-server/0 workload status is error since 2018-12-11 03:44:39Z
ERROR:root:rabbitmq-server/0 failed: workload status is error
2018-12-10 20:45:05,358 [DEBUG] conjure-up/openstack-novalxd - events.py:52 - Setting Error at conjureup/events.py:149
2018-12-10 20:45:05,360 [ERROR] conjure-up/openstack-novalxd - events.py:161 - Unhandled exception in <Task finished coro=<DeployController._wait_for_applications() done, defined at /snap/conjure-up/1034/lib/python3.6/site-packages/conjureup/controllers/juju/deploy/gui.py:81> exception=DeploymentFailure('Some applications failed to start successfully.',)>
Traceback (most recent call last):
File "/snap/conjure-up/1034/lib/python3.6/site-packages/conjureup/controllers/juju/deploy/gui.py", line 82, in _wait_for_applications
await common.wait_for_applications(view.set_footer)
File "/snap/conjure-up/1034/lib/python3.6/site-packages/conjureup/controllers/juju/deploy/common.py", line 50, in wait_for_applications
await juju.wait_for_deployment()
File "/snap/conjure-up/1034/lib/python3.6/site-packages/conjureup/juju.py", line 747, in wait_for_deployment
"Some applications failed to start successfully.")
conjureup.errors.DeploymentFailure: Some applications failed to start successfully.
2018-12-10 20:45:05,456 [DEBUG] conjure-up/openstack-novalxd - __init__.py:27 - Showing dialog for exception: Some applications failed to start successfully.第二行显示(安全的猜测)一个小时后的超时(加上服务器负载很重的几秒钟)(我是GMT-7),到目前为止已经复制了两次。
此时,Ceph集群“几乎”上升,其他东西被阻塞,等待日志中您看到的服务的另一半,这些服务还没有完全完成(它们仍然在安装软件包)。我想如果我们能等一个多小时就行了?
(rabbitmq已启动并运行,但由于服务器负载,它和其他服务在错误和就绪/等待之间波动,但很快就会自我纠正-即不应该触发此错误)。
是否可以通过命令行或其他方式配置超时?
服务器正在运行来自snap的Ubuntu18.04和LXD 3.7。lxd的默认存储设置为BTRFS,网络是一个本机桥(br0 -而不是lxd的lxdbr0)
发布于 2018-12-14 04:40:42
是的,切换到ZFS池完成了这个任务。部署工作用了5.8小时,员额部署又用了半小时。
我注意到的一个不同之处是内存利用率(我在另一个窗口中观看top )。随着BTRFS的波动,但很少超过3GB,而buff/缓存垄断了我的其余的内存。使用ZFS,它很快就达到了3-4GB+,并且继续攀升,而buff/cache保持了1GB的性能,直到部署结束时服务才开始运行。
因此,我认为使用BTRFS的fs缓存会给系统内存带来太大的压力,从而使进程无法在峰值效率下运行。
https://unix.stackexchange.com/questions/487267
复制相似问题