You are viewing a plain text version of this content. The canonical link for it is here.
Posted to users@cloudstack.apache.org by James Kahn <jk...@idea11.com.au> on 2012/07/15 12:16:08 UTC
XenServer and CloudStack - HA
A while ago we tested XenServer and CloudStack HA. This was with XS 6.0 and CS 3.0.
Initially with XenServer HA disabled and CloudStack HA enabled, we found that if the pool master dies, the VMs running on that host are not restarted on other hosts by CloudStack HA – at least for the 20 minutes we waited for something to happen. Not good. If any other host dies the VMs are restarted promptly without issue. We assumed that this is because CloudStack needs the pool master up.
Testing again with XS HA in use to protect the pool master, and CS HA enabled, HA functionality works. So we're running with XS HA protecting the pool master and CS HA protecting the VMs.
We've logged a Citrix case for an XS related issue and were told by the support engineer that his cloudstack support manager said "with cloudstack you should not use XS HA. Period. CS has got its own HA". Which would be fine, except that CS HA doesn't work if the pool master dies. At least with our testing.
What's the go here? Does CS HA now protect the pool master? How should we be configuring this?
Thanks,
JK.
答复: 答复: XenServer and CloudStack - HA
Posted by Mice Xia <mi...@tcloudcomputing.com>.
James,
There is a global configuration 'restart.retry.interval' that controls the retry interval of HA thread, can you try again with a smaller value, such as 120 or 60?
Regards
Mice
-----邮件原件-----
发件人: James Kahn [mailto:jkahn@idea11.com.au]
发送时间: 2012年7月16日 10:38
收件人: cloudstack-users@incubator.apache.org
主题: Re: 答复: XenServer and CloudStack - HA
Hi Mice,
When we tested we pulled the power from the pool master to simulate a
hardware failure.
Cheers,
JK.
-----Original Message-----
From: Mice Xia <mi...@tcloudcomputing.com>
Reply-To: "cloudstack-users@incubator.apache.org"
<cl...@incubator.apache.org>
Date: Monday, 16 July 2012 12:22 PM
To: "cloudstack-users@incubator.apache.org"
<cl...@incubator.apache.org>
Subject: 答复: XenServer and CloudStack - HA
>Hi James,
>
>I think CS HA does protect pool master. In your environment, did HA fail
>every time, or occasionally?
>And the way pool master die make some difference, was it a management
>network partition, a public network partition, a storage network
>partition, or a host power down?
>
>Regards
>Mice
>
>-----邮件原件-----
>发件人: James Kahn [mailto:jkahn@idea11.com.au]
>发送时间: 2012年7月15日 18:16
>收件人: cloudstack-users@incubator.apache.org
>主题: XenServer and CloudStack - HA
>
>A while ago we tested XenServer and CloudStack HA. This was with XS 6.0
>and CS 3.0.
>
>Initially with XenServer HA disabled and CloudStack HA enabled, we found
>that if the pool master dies, the VMs running on that host are not
>restarted on other hosts by CloudStack HA – at least for the 20 minutes
>we waited for something to happen. Not good. If any other host dies the
>VMs are restarted promptly without issue. We assumed that this is because
>CloudStack needs the pool master up.
>
>Testing again with XS HA in use to protect the pool master, and CS HA
>enabled, HA functionality works. So we're running with XS HA protecting
>the pool master and CS HA protecting the VMs.
>
>We've logged a Citrix case for an XS related issue and were told by the
>support engineer that his cloudstack support manager said "with
>cloudstack you should not use XS HA. Period. CS has got its own HA".
>Which would be fine, except that CS HA doesn't work if the pool master
>dies. At least with our testing.
>
>What's the go here? Does CS HA now protect the pool master? How should we
>be configuring this?
>
>Thanks,
>JK.
>
>
>
Re: 答复: XenServer and CloudStack - HA
Posted by James Kahn <jk...@idea11.com.au>.
Hi Mice,
When we tested we pulled the power from the pool master to simulate a
hardware failure.
Cheers,
JK.
-----Original Message-----
From: Mice Xia <mi...@tcloudcomputing.com>
Reply-To: "cloudstack-users@incubator.apache.org"
<cl...@incubator.apache.org>
Date: Monday, 16 July 2012 12:22 PM
To: "cloudstack-users@incubator.apache.org"
<cl...@incubator.apache.org>
Subject: 答复: XenServer and CloudStack - HA
>Hi James,
>
>I think CS HA does protect pool master. In your environment, did HA fail
>every time, or occasionally?
>And the way pool master die make some difference, was it a management
>network partition, a public network partition, a storage network
>partition, or a host power down?
>
>Regards
>Mice
>
>-----邮件原件-----
>发件人: James Kahn [mailto:jkahn@idea11.com.au]
>发送时间: 2012年7月15日 18:16
>收件人: cloudstack-users@incubator.apache.org
>主题: XenServer and CloudStack - HA
>
>A while ago we tested XenServer and CloudStack HA. This was with XS 6.0
>and CS 3.0.
>
>Initially with XenServer HA disabled and CloudStack HA enabled, we found
>that if the pool master dies, the VMs running on that host are not
>restarted on other hosts by CloudStack HA – at least for the 20 minutes
>we waited for something to happen. Not good. If any other host dies the
>VMs are restarted promptly without issue. We assumed that this is because
>CloudStack needs the pool master up.
>
>Testing again with XS HA in use to protect the pool master, and CS HA
>enabled, HA functionality works. So we're running with XS HA protecting
>the pool master and CS HA protecting the VMs.
>
>We've logged a Citrix case for an XS related issue and were told by the
>support engineer that his cloudstack support manager said "with
>cloudstack you should not use XS HA. Period. CS has got its own HA".
>Which would be fine, except that CS HA doesn't work if the pool master
>dies. At least with our testing.
>
>What's the go here? Does CS HA now protect the pool master? How should we
>be configuring this?
>
>Thanks,
>JK.
>
>
>
答复: XenServer and CloudStack - HA
Posted by Mice Xia <mi...@tcloudcomputing.com>.
Hi James,
I think CS HA does protect pool master. In your environment, did HA fail every time, or occasionally?
And the way pool master die make some difference, was it a management network partition, a public network partition, a storage network partition, or a host power down?
Regards
Mice
-----�ʼ�ԭ��-----
������: James Kahn [mailto:jkahn@idea11.com.au]
����ʱ��: 2012��7��15�� 18:16
�ռ���: cloudstack-users@incubator.apache.org
����: XenServer and CloudStack - HA
A while ago we tested XenServer and CloudStack HA. This was with XS 6.0 and CS 3.0.
Initially with XenServer HA disabled and CloudStack HA enabled, we found that if the pool master dies, the VMs running on that host are not restarted on other hosts by CloudStack HA �C at least for the 20 minutes we waited for something to happen. Not good. If any other host dies the VMs are restarted promptly without issue. We assumed that this is because CloudStack needs the pool master up.
Testing again with XS HA in use to protect the pool master, and CS HA enabled, HA functionality works. So we're running with XS HA protecting the pool master and CS HA protecting the VMs.
We've logged a Citrix case for an XS related issue and were told by the support engineer that his cloudstack support manager said "with cloudstack you should not use XS HA. Period. CS has got its own HA". Which would be fine, except that CS HA doesn't work if the pool master dies. At least with our testing.
What's the go here? Does CS HA now protect the pool master? How should we be configuring this?
Thanks,
JK.
Re: XenServer and CloudStack - HA
Posted by James Kahn <jk...@idea11.com.au>.
Ah, we've found the issue. We ran our testing with only two XenServer
hosts. XenServer will not automatically transfer the pool master role if
you only run two XenServer hosts and lose the pool master.
So it sounds like the correct configuration is enable CloudStack HA,
disable XenServer HA.
Cheers,
JK.
-----Original Message-----
From: James Kahn <jk...@idea11.com.au>
Reply-To: "cloudstack-users@incubator.apache.org"
<cl...@incubator.apache.org>
Date: Sunday, 15 July 2012 8:16 PM
To: "cloudstack-users@incubator.apache.org"
<cl...@incubator.apache.org>
Subject: XenServer and CloudStack - HA
>A while ago we tested XenServer and CloudStack HA. This was with XS 6.0
>and CS 3.0.
>
>Initially with XenServer HA disabled and CloudStack HA enabled, we found
>that if the pool master dies, the VMs running on that host are not
>restarted on other hosts by CloudStack HA at least for the 20 minutes
>we waited for something to happen. Not good. If any other host dies the
>VMs are restarted promptly without issue. We assumed that this is because
>CloudStack needs the pool master up.
>
>Testing again with XS HA in use to protect the pool master, and CS HA
>enabled, HA functionality works. So we're running with XS HA protecting
>the pool master and CS HA protecting the VMs.
>
>We've logged a Citrix case for an XS related issue and were told by the
>support engineer that his cloudstack support manager said "with
>cloudstack you should not use XS HA. Period. CS has got its own HA".
>Which would be fine, except that CS HA doesn't work if the pool master
>dies. At least with our testing.
>
>What's the go here? Does CS HA now protect the pool master? How should we
>be configuring this?
>
>Thanks,
>JK.
>
>