openstack team mailing list archive

[Chris Behrens <cbehrens@xxxxxxxxxxxx>]

I wonder if the issue is gone as well.

After the db_pool code was removed, I was doing some separate testing of sqlalchemy's with_lockmode()..  and I was able to reproduce a traceback from sqlalchemy that looked extremely similar or identical to the traceback we were hitting when using eventlet's db_pool.  This was many months ago.  I moved on to other things and I didn't keep any notes on what I was doing.  A couple months ago.. I decided I should go look at it again.  And I could no longer reproduce the issue... doing what I believe was the same test I was doing initially.

So...  I wasn't testing db_pool, however I thought I was seeing the same traceback.  And then I couldn't reproduce it anymore.  So, I wonder if these were the same issue... and that the issue is gone now.

I say it's worth trying the db_pool code again... but I'd make it an option so that we can enable/disable it.

- Chris

On Mar 6, 2012, at 9:39 AM, Yuriy Taraday wrote:

> Now I had no luck trying to reproduce problem that appeared with
> db_pool. Maybe it's gone?
> Who can walk me through the way to that bug? I think, we should fix it
> if it's not fixed already, bring back db_pool and then eventlet will
> be all good again, am I right?
> 
> Kind regards, Yuriy.
> 
> 
> 
> On Fri, Mar 2, 2012 at 04:57, Vishvananda Ishaya <vishvananda@xxxxxxxxx> wrote:
>> I agree.  It would be awesome if someone could actually make it work.  We
>> had a totally broken version using the eventlet db pool 6 months ago.
>> 
>> Vish
>> 
>> On Mar 1, 2012, at 4:20 PM, Joshua Harlow wrote:
>> 
>> Sad, especially since so much is using the database :-(
>> 
>> On 3/1/12 2:43 PM, "Adam Young" <ayoung@xxxxxxxxxx> wrote:
>> 
>> On 03/01/2012 02:48 PM, Vishvananda Ishaya wrote:
>>> On Mar 1, 2012, at 9:39 AM, Adam Young wrote:
>>> 
>>>> What would the drawbacks be? Probably the first thing people would look
>>>> to from Eventlet is performance. I don't have the hard numbers to compare
>>>> Eventlet to Apache HTTPD, but I do know that Apache is used in enough high
>>>> volume sites that I would not be overly concerned. The traffic in an
>>>> Openstack deployment to a Keystone server is going to be about two orders of
>>>> magnitude less than any other traffic, and is highly unlikely to be the
>>>> bottleneck.
>>> How did you arrive at this number? Every user has to hit keystone before
>>> making a request to any other service (unless they already have a token) and
>>> each service needs to authenticate that token. Any request that hits
>>> multiple services will hit keystone multiple times.  Without caching,
>>> keystone is by far the busiest service in an openstack install. Caching
>>> should fix some of this, but I don't know that I would expect it to be two
>>> orders of magnitude less.
>>> 
>>> Vish
>> 
>> 
>> Seeing as the SQL Alchemy code is blocking on each request,  I suspect
>> that performance is now soundly *not* a reason to want to stick with
>> eventlet.  My statement  that Eventlet is performant is based on the
>> assumption that the benefits of using Greenthreads are realized.  It
>> looks like that is not the case.
>> 
>> _______________________________________________
>> Mailing list: https://launchpad.net/~openstack
>> Post to     : openstack@xxxxxxxxxxxxxxxxxxx
>> Unsubscribe : https://launchpad.net/~openstack
>> More help   : https://help.launchpad.net/ListHelp
>> 
>> 
>> 
>> _______________________________________________
>> Mailing list: https://launchpad.net/~openstack
>> Post to     : openstack@xxxxxxxxxxxxxxxxxxx
>> Unsubscribe : https://launchpad.net/~openstack
>> More help   : https://help.launchpad.net/ListHelp
>> 
> 
> _______________________________________________
> Mailing list: https://launchpad.net/~openstack
> Post to     : openstack@xxxxxxxxxxxxxxxxxxx
> Unsubscribe : https://launchpad.net/~openstack
> More help   : https://help.launchpad.net/ListHelp