On Thursday, June 09, 2016 03:47:17 PM Nicolas Ecarnot wrote:
Le 09/06/2016 15:31, Alexander Wels a écrit :
> On Thursday, June 09, 2016 03:11:07 PM Nicolas Ecarnot wrote:
>> Le 09/06/2016 08:58, Anantha Raghava a écrit :
>>> Hello Alexander,
>>>
>>> You are right. The FQDN was the culprit. Once I entered the FQDN in
>>> /etc/hosts file, the problem got resolved.
>>
>> Hello,
>>
>> I have not understood clearly whether your were using hosted engine or
>> bare-metal?
>>
>> I'm asking that because - as Alex knows it - I'm facing this terribly
>> slowness since months, and we took a lot a time trying to figure it out,
>> to no avail.
>> We're using 3.6.5 and bare-metal engine.
>> I triple checked DNS resolution, added (anyway) relevant fqdn in
>> /etc/hosts, install (anyway) haveged and rebooted.
>> We checked all that with Alex, but seeing nothing obvious.
>>
>> So far, the web gui usage is still a real daily pain.
>
> And we verified the ping between your browser machine and the engine was
> low. The load on the engine was near 0. That your browser was either
> Chrome/FF and that the machine running the browser is plenty powerful.
>
> Did you ever disable the repos besides the 3.6 one and run a yum update
> (on
> the engine)?
Yes, I did.
Two DC are still running on 3.6.3, and four others are running on 3.6.5.
Amongst the 3.6.5, some were updated from below, and some were installed
from scratch (so no negative legacy nor prehistoric RPMs or repos).
> Greg one of the other maintainers recently did some performance
> improvements, I just don't remember which version they go into. either
> 3.6.5 or 3.6.6.
May you Cc: to Greg?
I asked him, he said 3.6.5 and 4.0. So if your 3.6.5 is the same then it
didn't solve your issue obviously. I am completely out of things to try or see
that could possibly be the problem.
> If I remember correctly for some reason it takes a long time to
get
> data from the engine to your browser, we just don't know why it takes so
> long as we profiled the engine running in your environment and that was
> all fine as well.
I recall correctly.
I'm also witnessing memory leaks, that I think were fixed recently.