tag:support.appharbor.com,2010-11-23:/discussions/problems/63737-chocolatey-moving-very-slowly-since-latest-deployAppHarbor: Discussion 2018-10-19T03:07:11Ztag:support.appharbor.com,2010-11-23:Comment/344657112014-09-05T14:34:26Z2014-09-05T14:34:26ZChocolatey moving very slowly since latest deploy.<div><p>FYI, this was the move to the new servers last night.</p></div>ferventcodertag:support.appharbor.com,2010-11-23:Comment/344657112014-09-05T14:39:52Z2014-09-05T14:39:52ZChocolatey moving very slowly since latest deploy.<div><p>Note this discussion - <a href=
"https://github.com/chocolatey/chocolatey.org/issues/91">https://github.com/chocolatey/chocolatey.org/issues/91</a></p></div>ferventcodertag:support.appharbor.com,2010-11-23:Comment/344657112014-09-05T14:55:07Z2014-09-05T14:55:07ZChocolatey moving very slowly since latest deploy.<div><p>Okay, something just started snapping...</p></div>ferventcodertag:support.appharbor.com,2010-11-23:Comment/344657112014-09-05T14:56:56Z2014-09-05T14:56:56ZChocolatey moving very slowly since latest deploy.<div><p>Can anyone give me an indication of what was happening over
there?</p></div>ferventcodertag:support.appharbor.com,2010-11-23:Comment/344657112014-09-05T18:26:04Z2014-09-05T18:31:36ZChocolatey moving very slowly since latest deploy.<div><p>Hey Rob,</p>
<p>It looks a bit odd - have you seen the Librato dashboard? It
appears that immediately following the deployment last night all
traffic was routed to only one worker. Currently looking into
whether that could be caused by the load balancer or a similar
platform-related issue.</p>
<p>The worker that was serving traffic was able to handle the load
for the first 4-5 hours after the deployment last night, but around
00:00-00:15AM it was maxing out the available resources and
requests started being queued. This lead to the slow response time
you and other Chocolatey users experienced.</p>
<p>The issue also appears to have been resolved immediately
following the deployment this morning. I can see that response
times were still slower than usual, but definitely faster (10x)
than the response time prior to the deployment. Response times
normalized within 15 minutes.</p>
<p>New deployments effectively resets the application's running
state as well as the routing configuration, so the issue could
potentially be caused by either (or both) of those. To rule out
anything related to the application:</p>
<ul>
<li>What changes were in the commit last night (4:37PM) and this
morning (7:39AM)? It sounds like you changed map/reduce behavior to
address the performance issue, but is it likely that it could've
caused the behavior we're seeing here?</li>
<li>What do you mean that the app was moved to new servers last
night?</li>
</ul>
<p>By the way - feel free to use the word "urgent" in tickets when
you have an issue that needs immediate attention (and particularly
when it may be platform-related). It'll be marked as critical then
and processed faster.</p>
<p>Best,<br>
Rune</p></div>runetag:support.appharbor.com,2010-11-23:Comment/344657112014-09-08T16:22:18Z2014-09-08T16:22:18ZChocolatey moving very slowly since latest deploy.<div><blockquote>
<p>What do you mean that the app was moved to new servers last
night?</p>
</blockquote>
<p>By this I meant since we just switched over to the load balancer
and servers, the app had just deployed the first time to make this
possible.</p></div>ferventcodertag:support.appharbor.com,2010-11-23:Comment/344657112014-09-08T16:24:30Z2014-09-08T16:24:30ZChocolatey moving very slowly since latest deploy.<div><blockquote>
<p>What changes were in the commit last night (4:37PM) and this
morning (7:39AM)? It sounds like you changed map/reduce behavior to
address the performance issue, but is it likely that it could've
caused the behavior we're seeing here?</p>
</blockquote>
<p>It's very likely to have caused the issue when working with a
load balancer. I was hoping to get more insight to something on the
back end that may have been responsible though. The only change
between those two times was to unhook the module from the
pipeline.</p></div>ferventcodertag:support.appharbor.com,2010-11-23:Comment/344657112014-09-08T21:31:14Z2014-09-08T21:31:14ZChocolatey moving very slowly since latest deploy.<div><p>Hi Rob,</p>
<p>Actually looked further into this and there is a potential issue
that can crop up if your load balancer uses WebSockets/very long
running request. It's almost certainly not an issue with your
application and a solution to address the load balancer issue is in
the works.</p>
<p>Does your application use WebSockets or long-running/polling
requests? If so this is the most likely explanation and will be
address by the solution that is being prepared. In the meanwhile we
can tweak the load balancer (disable WebSockets and reduce
keepalive) - it's unlikely to crop up again, but it might be a
better option for the short term.</p>
<p>Best,<br>
Rune</p></div>runetag:support.appharbor.com,2010-11-23:Comment/344657112014-10-28T20:21:34Z2014-10-28T20:21:34ZChocolatey moving very slowly since latest deploy.<div><p>This have resolved themselves... sorry I forgot to reply.</p></div>ferventcoder