On Fri, Oct 7, 2016 at 4:02 PM, Michal Skrivanek <michal.skrivanek@redhat.com> wrote:

On 7 Oct 2016, at 15:28, Simone Tiraboschi <stirabos@redhat.com> wrote:



On Fri, Oct 7, 2016 at 3:25 PM, Michal Skrivanek <michal.skrivanek@redhat.com> wrote:

On 7 Oct 2016, at 14:59, Nir Soffer <nsoffer@redhat.com> wrote:

On Fri, Oct 7, 2016 at 3:52 PM, Michal Skrivanek <michal.skrivanek@redhat.com> wrote:

On 7 Oct 2016, at 14:42, Nir Soffer <nsoffer@redhat.com> wrote:

On Wed, Oct 5, 2016 at 1:33 PM, Simone Tiraboschi <stirabos@redhat.com> wrote:


On Wed, Oct 5, 2016 at 10:34 AM, Nir Soffer <nsoffer@redhat.com> wrote:
On Wed, Oct 5, 2016 at 10:24 AM, Simone Tiraboschi <stirabos@redhat.com> wrote:


On Wed, Oct 5, 2016 at 9:17 AM, gregor <gregor_forum@catrix.at> wrote:
Hi,

did you found a solution or cause for this high CPU usage?
I have installed the self hosted engine on another server and there is
no VM running but ovirt-ha-agent uses heavily the CPU.

Yes, it's due to the fact that ovirt-ha-agent periodically reconnects over json rpc and this is CPU intensive since the client has to parse the yaml API specification each time it connects.

wasn’t it suppose to be fixed to reuse the connection? Like all the other clients (vdsm migration code:-) 

This is orthogonal issue.

Yes it is. And that’s the issue;-)
Both are wrong, but by “fixing” the schema validation only you lose the motivation to fix the meaningless wasteful reconnect

Yes, we are going to fix that too ( https://bugzilla.redhat.com/show_bug.cgi?id=1349829 )

that’s great! Also al the other vdsClient uses?:-)

https://gerrit.ovirt.org/#/c/62729/
 
What is that periodic one call anyway? Is there only one? Maybe we don’t need it so much.

Currently ovirt-ha-agent is periodically reconnecting the hosted-engine storage domain and checking its status. This is already on jsonrpc.
In 4.1 all the monitoring will be moved to jsonrpc.
 

but it would require also https://bugzilla.redhat.com/show_bug.cgi?id=1376843 to be fixed.

This is less good. Well, worst case you can reconnect yourself, all you need is a notification when the existing connection breaks

 

 
Does schema validation matter then if there would be only one connection at the start up?

Loading once does not help command line tools like vdsClient, hosted-engine and
vdsm-tool. 

none of the other tools is using json-rpc.

hosted-engine-setup is, and sooner or later we'll have to migrate also the remaining tools since xmlrpc has been deprecated with 4.0

ok. though setup is a one-time action so it’s not an issue there

 


Nir
 


Simone, reusing the connection is good idea anyway, but what you describe is 
a bug in the client library. The library does *not* need to load and parse the
schema at all for sending requests to vdsm.

The schema is only needed if you want to verify request parameters,
or provide online help, these are not needed in a client library.

Please file an infra bug about it.


Here is a patch that should eliminate most most of the problem:

Would be nice if it can be tested on the system showing this problem.

Cheers,
Nir
_______________________________________________
Users mailing list
Users@ovirt.org
http://lists.ovirt.org/mailman/listinfo/users


_______________________________________________
Users mailing list
Users@ovirt.org
http://lists.ovirt.org/mailman/listinfo/users


_______________________________________________
Users mailing list
Users@ovirt.org
http://lists.ovirt.org/mailman/listinfo/users


_______________________________________________
Users mailing list
Users@ovirt.org
http://lists.ovirt.org/mailman/listinfo/users