[ovirt-users] oVirt 4.0.x - hosted-engine was not starting properly

Thu Sep 29 13:11:43 UTC 2016

Hi Simone,

Thanks for the info. I'll look at the solution that you suggested.

Cheers,
Gervais

> On Sep 29, 2016, at 10:01 AM, Simone Tiraboschi <stirabos at redhat.com> wrote:
> 
> 
> 
> On Thu, Sep 29, 2016 at 2:51 PM, Gervais de Montbrun <gervais at demontbrun.com <mailto:gervais at demontbrun.com>> wrote:
> Hi Martin,
> 
> The entropy was super low. Somewhere around 140. I installed and configured haveged.service to start at bootup, reverted my apache changes... After a reboot, my systemctl status still says that there are 7 services queued (note that I erroneously said degraded in my previous email - the services are, in fact, queued), but the oVirt GUI comes up almost immediately and everything seems to be great.
> 
> 
> Take care that using havaged on a VM should not be considered a good source of entropy and the oVirt PKi is managed by the engine.
> http://security.stackexchange.com/questions/34523/is-it-appropriate-to-use-haveged-as-a-source-of-entropy-on-virtual-machines <http://security.stackexchange.com/questions/34523/is-it-appropriate-to-use-haveged-as-a-source-of-entropy-on-virtual-machines>
> 
> A better approach is the virtio-rng paravirtualised rng driver as for patch https://gerrit.ovirt.org/#/c/62334/ <https://gerrit.ovirt.org/#/c/62334/>
> 
>  
> Thank you for the tip. You solved my issue.
> 
> Cheers,
> Gervais
> 
> 
> 
>> On Sep 29, 2016, at 7:47 AM, Martin Perina <mperina at redhat.com <mailto:mperina at redhat.com>> wrote:
>> 
>> Hi,
>> 
>> please take a look at my inline comments:
>> 
>> On Tue, Sep 27, 2016 at 7:23 PM, Gervais de Montbrun <gervais at demontbrun.com <mailto:gervais at demontbrun.com>> wrote:
>> Hey All,
>> 
>> Since updating to 4.0.x of oVirt, I have had an issue with my hosted engine. After a some poking around, I think I have figured out my issue and thought I would share to see what others think.
>> The issue has existed with 4.0, 4.0.1, 4.0.2, 4.0.3, and still exists in 4.0.4.
>> 
>> Description:
>> When my hosted engine starts it reports that it is in a degraded state with 7 or 8 services still not started when I run systemctl status. It takes about 6 or 7 minutes to eventually start all the services and come online. If I don't set my cluster to Global-Maintenance mode it eventually thinks that my hosted-engine needs to be rebooted and restarts it before it can start everything.
>> 
>> Could you please share with us logs gathered by ovirt-log-collector?
>> 
>> It's just a guess but could you please take a look if you HE VM has enough entropy?
>> 
>>   cat /proc/sys/kernel/random/entropy_avail
>> 
>> If the value is low (below or around 200),  you really need to install and configure some entropy generator such as haveged
>> 
>> 
>> Solution:
>> I realized that Apache was the culprit and found that the proxy to the ovirt-engine in /etc/httpd/conf.d/z-ovirt-engine-proxy.conf has a super long timeout with many retries. I changed the settings and now everything works for me.
>> 
>> -> Before change:
>>     <LocationMatch ^/(ovirt-engine($|/)|api($|/)|RHEVManagerWeb/|OvirtEngineWeb/|ca.crt$|engine.ssh.key.txt$|rhevm.ssh.key.txt$)>
>>         ProxyPassMatch ajp://127.0.0.1:8702 <> timeout=3600 retry=5
>> 
>>         <IfModule deflate_module>
>>             AddOutputFilterByType DEFLATE text/javascript text/css text/html text/xml text/json application/xml application/json application/x-yaml
>>         </IfModule>
>>     </LocationMatch>
>> 
>> -> After change:
>>     <LocationMatch ^/ovirt-engine($|/)>
>>         ProxyPassMatch ajp://127.0.0.1:8702 <> timeout=5 retry=2
>> 
>>         <IfModule deflate_module>
>>             AddOutputFilterByType DEFLATE text/javascript text/css text/html text/xml text/json application/xml application/json application/x-yaml
>>         </IfModule>
>>     </LocationMatch>
>> 
>> This one is correct for 4.0, not sure why it was not updated during upgrade from 3.6. @Simone?
>>  
>> 
>> If I read the timeout settings correctly, it will wait 60 minutes with 5 retries. 5 hours is way too long for my little server to hold onto all those apache processes.
>> The change I made allows for there to be an error, and also releases apache's hold on the process. Once everything is ready, apache is ready to serve requests and everything/everyone is happy. Before making the change, I just get a whitescreen in my browser and then nothing works until I restart Apache (or I end up in an endless loop of ovirt-ha services restarting my hosted-engine.
>> 
>> Well, if you have an issue with too many apache processes waiting for engine to respond, then there's some issue in engine. As I wrote above please share the logs with us and check entropy.
>> 
>> Thanks
>> 
>> Martin Perina
>>  
>> 
>> I noticed that this setting reverts to the original setting, so oVirt must be writing this file. Perhaps these number can be changed in oVirt? If not, I will just setup and ansible play to revert the settings with working values and restart apache on my engine.
>> :-)
>> 
>> Cheers,
>> Gervais
>> 
>> 
>> 
>> 
>> _______________________________________________
>> Users mailing list
>> Users at ovirt.org <mailto:Users at ovirt.org>
>> http://lists.ovirt.org/mailman/listinfo/users <http://lists.ovirt.org/mailman/listinfo/users>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.ovirt.org/pipermail/users/attachments/20160929/c8ecb052/attachment-0001.html>