Description of problem: in a single host (local storage) DC you get the following error message in webadmin: Power Management test failed for Host $TESTHOST. There is no other host in the data center that can be used to test the power management settings. Version-Release number of selected component (if applicable): This bug is present since 3.1 at least. I'm sorry for not reporting it sooner, I thought someone else will do.. How reproducible: always Steps to Reproduce: 1. wait some time, until ovirt decides to test power management 2. get the above message 3. Actual results: you get the message that power management test fails, because there is no other host Expected results: different options: 1. do no power management test at all if there is only 1 host 2. do no power management test in local storage DC 3. do power management test, but don't report any errors. I think option 1. is the best, because in the future the limitations for local storage only DCs will drop and this error will likely apply to all DCs with just one host. Additional info:
additional info: I didn't test this yet with other data center types with a single host, so this also might apply to nfs & gluster, etc. DCs
This is not related to the DC type. We are requiring a proxy Host to be available in order to send fencing commands to the target Host.\ This is actually functioning exactly as designed. I think that it is much friendly to let the user know that he can not TEST the fencing agent because he has no proxy Host instead of disabling the TEST button I recommend to CLOSE as NOTABUG
Eli - the design does not allow for a local storage host (single host in cluster/dc) to ever be fenced. why not remove the limitation of the fencing host must be in same DC if user so chooses?
(In reply to Itamar Heim from comment #3) > Eli - the design does not allow for a local storage host (single host in > cluster/dc) to ever be fenced. > why not remove the limitation of the fencing host must be in same DC if user > so chooses? this is a different FRE - to extend the proxy selection to be system wide. But this is not the issue we are talking about here. The described behavior is the designed one, in addition all-in-one is a specific case where fencing is not really required. Sven - what is the use case you were aiming for ?
Well I want to utilize the supported local storage data center type. And it is nowhere mentioned, that you need a proxy host in the first place. Or please point me to some documentation where this is stated. So the use-case would be to use power management in the future maybe over rest api. Why can't the engine-host be the proxy if there is no other? This was what I expected the engine does, talk directly to the ipmi-interface. I configured the ipmi settings in engine and engine should be capable of doing this. There is no technical limitation to just allow fencing with a second host in this data center. And if you say it "works as designed" I can just reply that it's poorly designed. If this is the design I would at least expect that power management is simply disabled in all single host DCs with a good explanation why it is disabled and not to just fail if you want to use it.
Additional question which comes to my mind: How can you re-enable a host in a single host data center when it was shutdown manually outside oVirt? Powermanagement via ovirt does not work. So you have to trigger this outside of oVirt. Also how would the workflow be to put a host into maintenance mode, e.g. if you need to reboot it?
(In reply to Sven Kieske from comment #5) > Well I want to utilize the supported local storage data center type. > > And it is nowhere mentioned, that you need a proxy host in the first place. > Or please point me to some documentation where this is stated. It is stated in 3.3 documentation section 6.6.2 > > So the use-case would be to use power management in the future maybe over > rest api. Fencing operations are already available via REST API api/hosts/<host_id>/fence > > Why can't the engine-host be the proxy if there is no other? > This was what I expected the engine does, talk directly to the > ipmi-interface. > I configured the ipmi settings in engine and engine should be capable of > doing > this. We have a RFE for that , AFAIK it will not get to 3.4 This is a little complex since engine calls VDSM fenceNode in order to perform fencing operations , so , we will have to implement a lightweight VDSM in the engine side for that > > There is no technical limitation to just allow fencing with a second host in > this data center. We have this feature for several years, we are committed to improve it and as I stated above there is already a RFE for letting the engine be the proxy. As any software, you are start with basic abilities and add more according to the input we get from customers > > And if you say it "works as designed" I can just reply that it's poorly > designed. > > If this is the design I would at least expect that power management is simply > disabled in all single host DCs with a good explanation why it is disabled > and > not to just fail if you want to use it. Again , I argue about that , I think that a message telling you that you have no proxy Host is much friendly than disabling the button , I see no "logical" place to put the explanation you are talking about either
(In reply to Sven Kieske from comment #6) > Additional question which comes to my mind: > > How can you re-enable a host in a single host data center when it was > shutdown > manually outside oVirt? Currently , the only way is to start it manually in the same way you had stopped it > > Powermanagement via ovirt does not work. So you have to trigger this outside > of oVirt. > > Also how would the workflow be to put a host into maintenance mode, e.g. if > you need to reboot it? You will have to reboot it manually as well
(In reply to Eli Mesika from comment #7) > (In reply to Sven Kieske from comment #5) > > Well I want to utilize the supported local storage data center type. > > > > And it is nowhere mentioned, that you need a proxy host in the first place. > > Or please point me to some documentation where this is stated. > > It is stated in 3.3 documentation section 6.6.2 Well, this is still beta documentation and this is available to the public since when? A few weeks? Do you tell me this feature is also just a few weeks old? > > > > > So the use-case would be to use power management in the future maybe over > > rest api. > > Fencing operations are already available via REST API > > api/hosts/<host_id>/fence > Thats cool, but as long as it doesn't work with a single host I can't use it. > > > > Why can't the engine-host be the proxy if there is no other? > > This was what I expected the engine does, talk directly to the > > ipmi-interface. > > I configured the ipmi settings in engine and engine should be capable of > > doing > > this. > > We have a RFE for that , AFAIK it will not get to 3.4 > This is a little complex since engine calls VDSM fenceNode in order to > perform fencing operations , so , we will have to implement a lightweight > VDSM in the engine side for that > > > > > There is no technical limitation to just allow fencing with a second host in > > this data center. > > We have this feature for several years, we are committed to improve it and > as I stated above there is already a RFE for letting the engine be the > proxy. > As any software, you are start with basic abilities and add more according > to the input we get from customers > > > > > And if you say it "works as designed" I can just reply that it's poorly > > designed. > > > > If this is the design I would at least expect that power management is simply > > disabled in all single host DCs with a good explanation why it is disabled > > and > > not to just fail if you want to use it. > > Again , I argue about that , I think that a message telling you that you > have no proxy Host is much friendly than disabling the button , I see no > "logical" place to put the explanation you are talking about either If a button is there and clickable I assume that it will work. Other features also get grayed out if they can not work in the current setup. Why would you like to be able to click a button which does not work? I would expect the button to be grayed out and when I hover over it, it should provide a tool tip saying e.g. "you can't use power management atm because there is no fencing host, please configure/install a fencing host first ". It would be cool if you could mail me the number of the RFE you are talking about (I hope it's visible to the public). Thanks for your time so far. We'll need to just use power management without ovirt until this is resolved. At least ovirt shouldn't complain about missing power management configuration. There should be a working state for a single host DC where you don't get error events: (option 1: I configure power management: ovirt complains that it can't test it, because there is no other host option 2: I do not configure power management: ovirt complains because power management is not configured).
This is an automated message. Re-targeting all non-blocker bugs still open on 3.4.0 to 3.4.1.
Closing old bugs. If this issue is still relevant/important in current version, please re-open the bug.