91
Bugs / Re: Installation of OMS fails
« Last post by Decoyed on February 28, 2013, 10:04:57 PM »Deemoowoor, thanks for hints!
best
Dec.
best
Dec.
Recent Posts
92
Bugs / Re: New VM not visible in OMS
« Last post by Decoyed on February 28, 2013, 10:04:24 PM »Ok, now I understand. And re templates, I've logged out/in, restarted etc. and still have only omv mv. But don't worry about it 
Once again thank you for your help and time,
Best
Dec.
Once again thank you for your help and time,
Best
Dec.
93
Bugs / Re: Installation of OMS fails
« Last post by deemoowoor on February 27, 2013, 02:01:28 PM »Hi,
Latest version of OMS now has logging level at INFO by default. You can tweak logging levels quite precisely by creating a file called "logging.conf", which follows the standard python logging module's format (see http://docs.python.org/2/library/logging.config.html#configuration-file-format):
This file's config overrides whatever is specified in opennode-oms.conf about logging.
Latest version of OMS now has logging level at INFO by default. You can tweak logging levels quite precisely by creating a file called "logging.conf", which follows the standard python logging module's format (see http://docs.python.org/2/library/logging.config.html#configuration-file-format):
Code: [Select]
[formatters]
keys=default
[formatter_default]
format=%(asctime)s %(thread)x %(name)s %(levelname)s %(message)s
[handlers]
keys=default
[handler_default]
class=logging.FileHandler
args=('omsd.log',)
formatter=default
[loggers]
keys=root
[logger_root]
level=INFO
handlers=default
This file's config overrides whatever is specified in opennode-oms.conf about logging.
94
Installation and Configuration / Re: Opennode HA Cluster Question
« Last post by activesys on February 27, 2013, 12:11:34 PM »1) HA failover will incur some downtime - in case of simple-HA its abou 1-2 minutes, in case of VM (advanced) HA its about 20sec (configurable somewhat - depends on heartbeat detection timeouts) - but also if some node hardware dies then full switchover can be around 1 minute or so - depends how many VM need to be restarted on another host.
There is currently no known "online" HA (with state preserving) solutions for QEMU/KVM or OpenVZ (Im excluding load-balance clusters atm) - primary reason being the perfomance degradation (overhead of continous VM checkpointing). We are investigating some possible future solutions here - but it takes time for these to mature.
2) RHCS cluster tools are mainly: clustat and clusvcadm - please read official RHCS docs for Cluster Administration:
https://access.redhat.com/knowledge/docs/en-US/Red_Hat_Enterprise_Linux/6/html/Cluster_Administration/index.html
You can use TUI for managing VMs safely on case of Simple-HA - as then cluster HA services are OpenVZ service and KVM service (not VMs itself).
3) Assuming here that you correctly configured cluster fencing - then you should test cluster failover by fencing another node like this: fence_node hostname
There is currently no known "online" HA (with state preserving) solutions for QEMU/KVM or OpenVZ (Im excluding load-balance clusters atm) - primary reason being the perfomance degradation (overhead of continous VM checkpointing). We are investigating some possible future solutions here - but it takes time for these to mature.
2) RHCS cluster tools are mainly: clustat and clusvcadm - please read official RHCS docs for Cluster Administration:
https://access.redhat.com/knowledge/docs/en-US/Red_Hat_Enterprise_Linux/6/html/Cluster_Administration/index.html
You can use TUI for managing VMs safely on case of Simple-HA - as then cluster HA services are OpenVZ service and KVM service (not VMs itself).
3) Assuming here that you correctly configured cluster fencing - then you should test cluster failover by fencing another node like this: fence_node hostname
95
Installation and Configuration / Re: Opennode HA Cluster Question
« Last post by fbmondia on February 27, 2013, 06:52:10 AM »Hi,
By the way, how do I test the 2-node HA? Is it just as simple as shutting-down the master node and expecting the VMs to be still online/restart on the second node?
Francis
By the way, how do I test the 2-node HA? Is it just as simple as shutting-down the master node and expecting the VMs to be still online/restart on the second node?
Francis
96
Installation and Configuration / Re: Opennode HA Cluster Question
« Last post by fbmondia on February 27, 2013, 06:42:24 AM »Hi,
A question on this.. Does this mean that the VMs will be re-started on the 2nd node and incur downtime? Or no downtime will be incurred during this switch-over?
Francis
Quote
simple HA approach where High-Availablity service will be the hypervisor service - not VM itself. So cluster checks hypervisor status (vz and kvm) - if that service or server hardware fails all VMs will be restarted on another server/cluster node.
A question on this.. Does this mean that the VMs will be re-started on the 2nd node and incur downtime? Or no downtime will be incurred during this switch-over?
Francis
97
Bugs / Re: New VM not visible in OMS
« Last post by ilja_l on February 26, 2013, 11:03:52 PM »Hi,
unfortunately the nature of container virtualization (with OpenVZ belonging to that level) doesn't allow to run OSs with a different kernel (http://en.wikipedia.org/wiki/Operating_system-level_virtualization), so it's only linux flavours with openvz. Windows/freebsd need to go to KVM.
2. Templates should be autodetected and populated as well, but it's not super dynamic at the moment - need to refresh the onc model, e.g. by logging out/in or refreshing. Going to change as it's pretty annoying
unfortunately the nature of container virtualization (with OpenVZ belonging to that level) doesn't allow to run OSs with a different kernel (http://en.wikipedia.org/wiki/Operating_system-level_virtualization), so it's only linux flavours with openvz. Windows/freebsd need to go to KVM.
2. Templates should be autodetected and populated as well, but it's not super dynamic at the moment - need to refresh the onc model, e.g. by logging out/in or refreshing. Going to change as it's pretty annoying
98
Bugs / Re: New VM not visible in OMS
« Last post by Decoyed on February 26, 2013, 10:50:30 PM »Ilya,
thanks for prompt answer. And sorry for stupid question, but is there an OpenVZ template for Windows-based VM and/or FreeBSD?
I've also downloaded some templates (ubuntu, fedora...) but when selecting "New" in OMS I can see only "Opennode OMS" template. Is it normal or I'm doing something wrong (as always).
Sorry for spamming the thread and forum, but I need to teach some people how to use it.
Thanks four your efforts and work,
Dec.
thanks for prompt answer. And sorry for stupid question, but is there an OpenVZ template for Windows-based VM and/or FreeBSD?
I've also downloaded some templates (ubuntu, fedora...) but when selecting "New" in OMS I can see only "Opennode OMS" template. Is it normal or I'm doing something wrong (as always).
Sorry for spamming the thread and forum, but I need to teach some people how to use it.
Thanks four your efforts and work,
Dec.
99
Bugs / Re: New VM not visible in OMS
« Last post by ilja_l on February 26, 2013, 08:31:46 PM »Hi,
oh yes, Ilja has an idea OMS at the moment only supports management of OpenVZ-based VMs. We started by supporting both via libvirt, unfortunately libvirt is too high level abstraction, so we've switched to hypervisor specific functionality starting from OpenVZ containers as this is what we use the most.
Most of the code is agnostic to the hypervisor, but e.g. ONC (web console) relies on certain data being available that we couldn't get easily from KVM. As we had priority on stabilisation, 'plugging in KVM back' was delayed. It's not the next most important task, but pretty close on the horizon.
cheers,
Ilya
oh yes, Ilja has an idea
OMS at the moment only supports management of OpenVZ-based VMs. We started by supporting both via libvirt, unfortunately libvirt is too high level abstraction, so we've switched to hypervisor specific functionality starting from OpenVZ containers as this is what we use the most. Most of the code is agnostic to the hypervisor, but e.g. ONC (web console) relies on certain data being available that we couldn't get easily from KVM. As we had priority on stabilisation, 'plugging in KVM back' was delayed. It's not the next most important task, but pretty close on the horizon.
cheers,
Ilya
100
Installation and Configuration / Re: Opennode HA Cluster Question
« Last post by fbmondia on February 26, 2013, 05:39:41 PM »Hi Again,
I have successfully configured my prototype system. Sent you an email about it.
Question:
1. What cluster tools can be used to manage VM state? For example, creating, modifying, changing states, etc.
2. How do I use clusvcadm tool to control VM state?
3. What about this then?
Francis
I have successfully configured my prototype system. Sent you an email about it.
Question:
Quote
please note that if VM is put under HA cluster service - then you MUST manage VM state (starting/stopping) also with cluster tools.
1. What cluster tools can be used to manage VM state? For example, creating, modifying, changing states, etc.
Quote
Currently TUI/OMS wont detect VMs under HA - and wont use cluster tools for such VM management. However they can be still used to deploy VMs into RHCS HA cluster and then configure HA service and control VM state with clusvcadm tool
2. How do I use clusvcadm tool to control VM state?
Quote
TUI offers a local management of VMs, which are running on the same host. So, if a system is configured for HA and the failover brings new VMs to a certain host, then yes, TUI will be able to show/manage those VMs.
3. What about this then?
Francis