[ovirt-users] fakevdsm templates per datacenter

[Roy Golan]

On 27 October 2016 at 10:42, <joost at familiealbers.nl> wrote:

> Hi All, i have two ovirt / fakevdsm questions.
>
> I am using fakevdsm to test my ovirt 4.0.5 engine.
> I have a somewhat awkward setup where i have a relatively large amount of
> dc's.

I cannot get nfs master domain to work from the api.
> The api code is solid and is used on a ovirt-engine 3.5 production cluster.
>
>

> When creating the nfs storage domain i have no issues (it is created as id
> (00000-00000 more zeros here......) but when i try to attach it to a dc it
> fails.
>

please share the logs


> I am not sure this is an ovirt of fakevdsm issue. When i manually add the
> nfs storage domain it does get created which is somewhat worrying.
>

Manually?


> My real aim would be to run 500 dc's with 2 hosts and around 5 vms.
>
>
You'd have to have a host per DC, cause every DC has an SPM. You meant
cluster or I'm missing something.


> Not being able to use the nfs storage master domain is one of the issues.
>
> To work around this i decided to try and create 1000 dcs with a single
> host using local storage.
> Allthough this works it looks like the hosts and vms within the dc's are
> actually not having their own storage pool.
>
> In our prod environment we run at a much lower scale serveral dc's with
> shared gluster storage (shared between the two hosts in the cluster)
> This is considered per dc its own storage pool.
>
> In my tests when reaching dc,cluster, host 249 things fail, this is likely
> because of the value of 'maxnumberofhostsinstoragepool'. I would expect
> this to be bound to the dcs and not the overall server.
>
> In short i expect each dc to be able to run 249 hosts as each dc is its
> own storage pool?
>
> I'm not sure I understand the question. However, the limitation is soft
and can be overridden.


> Similarly when i look at the ovirt ui some hosts actually show they are
> running the total amount of running vms.
>

Yes, when running with vdsm fake you must set this config value as well

```bash

sudo -i -u postgreexport ENGINE_DB=dbname
psql $ENGINE_DB -c "UPDATE vdc_options set option_value = 'true' WHERE
option_name = 'UseHostNameIdentifier';"

 ```

I've sent a patch to gerrit for that  https://gerrit.ovirt.org/65804


> Again i am not sure if i am correct in assuming each dc is its own storage
> pool.
>
> Correct. DC = Storage Pool

>
> Finally.
> As this setup is somewhat awkward i would like to tweak the diverse flags
> in ovirt engine to be more relaxed about obtaining data from the hosts and
> storage domains.
>
> I believe ovirt engine now checks every 3 seconds the host but i am
> guessing i can live with every 60 seconds for example, the same would apply
> for the storage part. what i am not sure about though is if there is a
> golden ratio between these config items.
> I.e. there is vdsRefreshRate, if i set this to 60 instead of 3 do i need
> to also change the vdsTimeout value which is set to 180 by default?
>
> Change the VdsRefreshRate to 20s - but bare in mind that we have a
limitation when starting up a VM - its status will show as up only after
the next 20s interval
Please report with your finding on the above tweak


> sorru for the lenght of this message.
> the most critical items are.
> Should i worry about not being eable to create an nfs master domain or is
> this a fakevdsm thing
> Why are my dcs, for someone reason related to the main storage pool or am
> i right in thinking a dc + cluster +host with local storage, is its own
> storage pool?
>
> Thanks you so much.
>
>
>
>
>
>
> _______________________________________________
> Users mailing list
> Users at ovirt.org
> http://lists.ovirt.org/mailman/listinfo/users
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.ovirt.org/pipermail/users/attachments/20161027/be3bd66c/attachment-0001.html>