<div dir="ltr">As an interim update, adding the recommended ActiveRecord::Base.connection_pool.disconnect! doesn't seem to help :(<div><br></div><div>We're seeing far fewer timeouts but they're still happening on that db connection. And when they happen, it's carnage for 15s.</div></div><div class="gmail_extra"><br clear="all"><div><div class="gmail_signature"><div dir="ltr"><div><div dir="ltr"><div><div dir="ltr"><div><br></div><div><br></div><div><div>Simon Morley<br><span style="font-family:arial;font-size:small"><br></span></div><div><span style="font-family:arial;font-size:small">Big Chief</span> | PolkaSpots Supafly Wi-Fi<div>Bigger Chief | Cucumber Tony</div><div><br><div>Got an unlicensed Meraki? Set it free with Cucumber</div></div></div></div><div><a href="http://cucumberwifi.io/meraki" target="_blank">cucumberwifi.io/meraki</a></div><div><br></div></div></div></div></div></div></div></div>
<br><div class="gmail_quote">On 18 February 2016 at 14:15, Riccardo Tacconi <span dir="ltr"><<a href="mailto:rtacconi@gmail.com" target="_blank">rtacconi@gmail.com</a>></span> wrote:<br><blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex"><div dir="ltr">Sorry I meant you could use background jobs to query the DB, puma could just create a job without connection to the DB, that each job will have a connection. The main problem seems that you are using one process (worker) and one thread (because MRI uses one thread), which is the most inefficient thing to do. I think if you will put the database pool to one it will not change anything, I meant it would not be worser than now. I would try to:<div><br></div><div><ul><li>use real-threads, rubinius or JRuby,</li><li>ActiveRecord::Base.connection.disconnect! before forking<br></li><li>ActiveRecord::Base.establish_connection after forking<br></li><li>use more than one puma worker, say 1 : 50 connections, you will find the right ratio</li><li>not really sure about the connection pool, may be should be equal to puma workers times puma threads (real threads)</li></ul><div>Your endpoint cannot connected and do what is supposed to do in time, that's way it does not have time to close connections. So you need to have more processing power, which comes from parallel processing, by using processes and / or threads, this power should be used, in part to connect to the DB, and having more connections will help too, but this is in the end of the line.</div></div><div><br></div><div>If you will use background job, you will have to think about how to send back the processing, but it is usually the most scalable option. However first you have you Ruby, Puma, DB config setup properly :-)</div></div><div class="HOEnZb"><div class="h5"><div class="gmail_extra"><br><div class="gmail_quote">On 18 February 2016 at 13:30, Simon Morley <span dir="ltr"><<a href="mailto:simon@polkaspots.com" target="_blank">simon@polkaspots.com</a>></span> wrote:<br><blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex"><div dir="ltr">I'm afraid I have no virtually no idea what you mean! :( <div><br></div><div><div>We can't swap from cruby to jruby - mainly because that's like inviting someone else to the miserable party we're already hosting.</div></div><div><br></div><div><br></div></div><div class="gmail_extra"><span><br clear="all"><div><div><div dir="ltr"><div><div dir="ltr"><div><div dir="ltr"><div><br></div><div><br></div><div><div>Simon Morley<br><span style="font-family:arial;font-size:small"><br></span></div><div><span style="font-family:arial;font-size:small">Big Chief</span> | PolkaSpots Supafly Wi-Fi<div>Bigger Chief | Cucumber Tony</div><div><br><div>Got an unlicensed Meraki? Set it free with Cucumber</div></div></div></div><div><a href="http://cucumberwifi.io/meraki" target="_blank">cucumberwifi.io/meraki</a></div><div><br></div></div></div></div></div></div></div></div>
<br></span><div><div><div class="gmail_quote">On 18 February 2016 at 12:43, Riccardo Tacconi <span dir="ltr"><<a href="mailto:rtacconi@gmail.com" target="_blank">rtacconi@gmail.com</a>></span> wrote:<br><blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex"><div dir="ltr">Ruby MRI? If yes what's the point of using Puma? With MRI you have one worker and one thread, which is very inefficient. Would be possible to split HTTP requests handling from querying the DB? From Puma you could send requests to a topic (MOM), and multiple workers could process requests and each worker will have a DB connection. This could work with MRI, although you will need more RAM. However I would try rubinius of jruby first.<div><br></div><div>Sorry if I misunderstood, I did not follow the whole thread.<div><div><br><div class="gmail_extra"><br><div class="gmail_quote">On 18 February 2016 at 12:28, Simon Morley <span dir="ltr"><<a href="mailto:simon@polkaspots.com" target="_blank">simon@polkaspots.com</a>></span> wrote:<br><blockquote class="gmail_quote" style="margin:0px 0px 0px 0.8ex;border-left-width:1px;border-left-color:rgb(204,204,204);border-left-style:solid;padding-left:1ex"><div dir="ltr">Ruby 2.2.2 <div>Rails 4.2.5.1</div><div>mysql2 0.4.2 (tried a few)</div><div>Puma 2.16.0</div></div><div class="gmail_extra"><span><br clear="all"><div><div><div dir="ltr"><div><div dir="ltr"><div><div dir="ltr"><div><br></div><div><br></div><div><div>Simon Morley<br><span style="font-family:arial;font-size:small"><br></span></div><div><span style="font-family:arial;font-size:small">Big Chief</span> | PolkaSpots Supafly Wi-Fi<div>Bigger Chief | Cucumber Tony</div><div><br><div>Got an unlicensed Meraki? Set it free with Cucumber</div></div></div></div><div><a href="http://cucumberwifi.io/meraki" target="_blank">cucumberwifi.io/meraki</a></div><div><br></div></div></div></div></div></div></div></div>
<br></span><div><div><div class="gmail_quote">On 18 February 2016 at 12:24, Riccardo Tacconi <span dir="ltr"><<a href="mailto:rtacconi@gmail.com" target="_blank">rtacconi@gmail.com</a>></span> wrote:<br><blockquote class="gmail_quote" style="margin:0px 0px 0px 0.8ex;border-left-width:1px;border-left-color:rgb(204,204,204);border-left-style:solid;padding-left:1ex"><div dir="ltr">Which version of Ruby are you using?</div><div class="gmail_extra"><div><div><br><div class="gmail_quote">On 18 February 2016 at 12:17, Simon Morley <span dir="ltr"><<a href="mailto:simon@polkaspots.com" target="_blank">simon@polkaspots.com</a>></span> wrote:<br><blockquote class="gmail_quote" style="margin:0px 0px 0px 0.8ex;border-left-width:1px;border-left-color:rgb(204,204,204);border-left-style:solid;padding-left:1ex"><div dir="ltr">Actually puma docs suggest doing that when using preload_app and ActiveRecord...<div><br></div><div><a href="https://github.com/puma/puma#clustered-mode" target="_blank">https://github.com/puma/puma#clustered-mode</a><br></div></div><div class="gmail_extra"><br clear="all"><div><div><div dir="ltr"><div><div dir="ltr"><div><div dir="ltr"><div><br></div><div><br></div><div><div>Simon Morley<br><span style="font-family:arial;font-size:small"><br></span></div><div><span><span style="font-family:arial;font-size:small">Big Chief</span> | PolkaSpots Supafly Wi-Fi<div>Bigger Chief | Cucumber Tony</div></span><div><br><span><div>Got an unlicensed Meraki? Set it free with Cucumber</div></span></div></div></div><div><a href="http://cucumberwifi.io/meraki" target="_blank">cucumberwifi.io/meraki</a></div><div><br></div></div></div></div></div></div></div></div>
<br><div class="gmail_quote"><span>On 18 February 2016 at 12:05, Frederick Cheung <span dir="ltr"><<a href="mailto:frederick.cheung@gmail.com" target="_blank">frederick.cheung@gmail.com</a>></span> wrote:<br></span><div><div><blockquote class="gmail_quote" style="margin:0px 0px 0px 0.8ex;border-left-width:1px;border-left-color:rgb(204,204,204);border-left-style:solid;padding-left:1ex"><div style="word-wrap:break-word"><span><div style="font-family:Helvetica,Arial;font-size:13px;color:rgb(0,0,0);margin:0px"><br><br></div> <br><p>On 18 February 2016 at 11:17:34, Simon Morley (<a href="mailto:simon@polkaspots.com" target="_blank">simon@polkaspots.com</a>) wrote:</p> <blockquote type="cite"><span><div><div></div><div>
<div dir="ltr">
<div><br></div><div><div class="gmail_extra"><div><div><div dir="ltr"><div><div dir="ltr"><div dir="ltr">
<div>class RadiusDatabase </div>
<div>
<div> self.abstract_class = true</div>
<div> establish_connection "radius_#{Rails.env}".to_sym</div>
</div>
<div>end</div>
<div><br></div>
<div>class Radacct < RadiusDatabase</div>
<div>end</div>
<div><br></div>
<div>Then I decreased our database pool from 20 to 5 and added a
wait_timeout of 5 (since there seems to be some discrepancies with
this). Things got much better (but weren't fixed).</div>
<div><br></div>
<div>I tried querying differently, including
using <span style="color:rgb(51,51,51);font-family:Consolas,'Liberation Mono',Menlo,Courier,monospace;font-size:12px;line-height:22.4px;white-space:pre-wrap">connection_pool.with_connection.</span>
I've tried closing the connections manually and also used
ActiveRecord::Base.clear_active_connections! periodically. No
joy.</div>
<div><br></div>
<div>By this point, we were running 2-4 instances - handling around
very little traffic in total (about 50rpm). Every few hours, they'd
block, all of them. At the same time, we'd see a load of rack
timeouts - same DB. I've checked the connections - they were each
opening only a few to MySQL and MySQL was looking good.</div>
<div><br></div>
<div>One day, by chance, I reduced the 4 instances to 1. <b>And the
problem is solved!!! WHAT</b>? Obviously the problem isn't solved,
we can only use a single server. </div>
<div><br></div></div></div></div></div></div></div></div></div></div></div></div></span></blockquote><div><br></div></span>Are you using puma in the mode where it forks workers? if so, then you want to reconnect post fork or multiple processes will share the same file descriptor and really weird shit will happen.<div><br></div><div>The puma readme advises to do this:</div><div><br></div><div><div>before_fork do</div><div> ActiveRecord::Base.connection_pool.disconnect!</div><div>end</div><div><br></div><div>I don't know off the top of my head whether that will do the job for classes that have established a connection to a different db - presumably they have a separate connection pool</div><div><br></div><div>Fred</div><div><div><div><br><blockquote type="cite"><span><div><div><div dir="ltr"><div><div class="gmail_extra"><div><div><div dir="ltr"><div><div dir="ltr"><div dir="ltr">
<div>I don't know what's going on here. Have I been staring at this
for too long (yes)?</div>
<div><br></div>
<div>
<div>Our other servers are chugging along happily now, using a
connection pool of 20, no errors, no timeouts (different db
though).</div>
</div>
<div><br></div>
<div>Has anyone got any suggestions / seen this? Is there something
fundamentally wrong with the way we're establishing a connection to
the external dbs? Surely this is MySQL related</div>
<div><br></div>
<div>Thanks for listening, </div>
<div><br></div>
<div>S</div>
<div><br></div>
<div><br></div>
<div>
<div>Simon Morley<br>
<span style="font-family:arial;font-size:small"><br></span></div>
<div>
<div>
<div>Got an unlicensed Meraki? Set it free with Cucumber</div>
</div>
</div>
</div>
<div><a href="http://cucumberwifi.io/meraki" target="_blank">cucumberwifi.io/meraki</a></div>
<div><br></div>
</div>
</div>
</div>
</div>
</div>
</div>
<br>
<div class="gmail_quote">On 15 January 2016 at 13:58, Gerhard Lazu
<span dir="ltr"><<a href="mailto:gerhard@lazu.co.uk" target="_blank">gerhard@lazu.co.uk</a>></span> wrote:<br>
<blockquote class="gmail_quote" style="margin:0px 0px 0px 0.8ex;border-left-width:1px;border-left-color:rgb(204,204,204);border-left-style:solid;padding-left:1ex">
<div dir="ltr">
<div style="font-family:arial,helvetica,sans-serif;font-size:small;color:rgb(0,0,0)">
The understanding of difficult problems/bugs and the learning that
comes with it cannot be rushed. Each and every one of us has his /
her own pace, and all "speeds" are perfectly fine. The only
question that really matters is whether it's worth it (a.k.a. the
cost of lost opportunity). If the answer is yes, plough on. If not,
look for alternatives.</div>
<div style="font-family:arial,helvetica,sans-serif;font-size:small;color:rgb(0,0,0)">
<br></div>
<div style="font-family:arial,helvetica,sans-serif;font-size:small;color:rgb(0,0,0)">
Not everyone likes or wants to run their own infrastructure. The
monthly savings on the PaaS, IaaS advertised costs are undisputed,
but few like to think - never mind talk - about how many hours /
days / weeks have been spent debugging obscure problems which
"solve themselves" on a managed environment. Don't get me started
on those that are building their own Docker-based PaaS-es without
even realising it...</div>
<div style="font-family:arial,helvetica,sans-serif;font-size:small;color:rgb(0,0,0)">
<br></div>
<div style="font-family:arial,helvetica,sans-serif;font-size:small;color:rgb(0,0,0)">
As a side-note, I've been dealing with a similar TCP-related
problem for a while now, so I could empathise with your struggles
the second I've seen your post. One of us is bound to solve it
first, and I hope it will be you ; )</div>
<div style="font-family:arial,helvetica,sans-serif;font-size:small;color:rgb(0,0,0)">
<br></div>
<div style="font-family:arial,helvetica,sans-serif;font-size:small;color:rgb(0,0,0)">
Have a good one, Gerhard.</div>
<div>
<div>
<div class="gmail_extra"><br>
<div class="gmail_quote">On Fri, Jan 15, 2016 at 10:01 AM, Simon
Morley <span dir="ltr"><<a href="mailto:simon@polkaspots.com" target="_blank">simon@polkaspots.com</a>></span> wrote:<br>
<blockquote class="gmail_quote" style="margin:0px 0px 0px 0.8ex;border-left-width:1px;border-left-color:rgb(204,204,204);border-left-style:solid;padding-left:1ex">
<div dir="ltr">You must be more patient that I am. It's been a long
month - having said that, I'm excited to find the cause.
<div><br></div>
<div>I misunderstood you re. file descriptors. We checked the
kernel limits / files open on the systems before and during and
there's nothing untoward. </div>
<div><br></div>
<div>Since writing in, it's not happened as before - no doubt it'll
take place during our forthcoming office move today.</div>
<div><br></div>
<div>I ran a strace (thanks for that suggestion John) on a couple
of processes yesterday and saw redis blocking. Restarted a few
redis servers to see if that helped. Can't be certain yet.</div>
<div><br></div>
<div>As soon as it's on, I'll run a tcpdump. How I'd not thought
about that I don't know...</div>
<div><br></div>
<div>Actually, this is one thing I dislike about Rails - it's so
nice and easy to do everything, one forgets we're dealing with the
real servers / components / connections. It's too abstract in ways,
but that's a whole other debate :)</div>
<div><span><font color="#888888"><br></font></span></div>
<div><span><font color="#888888">S</font></span></div>
</div>
<div class="gmail_extra"><span><br clear="all"></span>
<div>
<div>
<div dir="ltr">
<div>
<div dir="ltr">
<div><span><br></span></div>
<div><span><br></span></div>
<div>
<div><span>Simon Morley<br>
<span style="font-family:arial;font-size:small"><br>
</span></span></div>
<div><span style="font-family:arial;font-size:small">Big
Chief</span> | PolkaSpots Supafly Wi-Fi
<div>Bigger Chief | Cucumber Tony</div>
<div><br>
<div><span style="font-size:small;font-family:arial">simon@PolkaSpots.com</span><br style="font-size:small;font-family:arial">
<div style="font-size:small;font-family:arial">Linkedin: I'm on it
again and it still sucks</div>
<div style="font-size:small;font-family:arial"><a href="tel:020%207183%201471" value="+442071831471" target="_blank">020
7183 1471</a></div>
</div>
</div>
</div>
</div>
<div style="font-size:small;font-family:arial"><br></div>
<div><font face="arial" size="2">🚀💥</font><br></div>
</div>
</div>
</div>
</div>
</div>
<br>
<div>
<div>
<div class="gmail_quote">On 15 January 2016 at 06:53, Gerhard Lazu
<span dir="ltr"><<a href="mailto:gerhard@lazu.co.uk" target="_blank">gerhard@lazu.co.uk</a>></span> wrote:<br>
<blockquote class="gmail_quote" style="margin:0px 0px 0px 0.8ex;border-left-width:1px;border-left-color:rgb(204,204,204);border-left-style:solid;padding-left:1ex">
<div dir="ltr">
<div style="font-family:arial,helvetica,sans-serif;font-size:small;color:rgb(0,0,0)">
File descriptors, for traditional reasons, include TCP
connections.<br></div>
<div style="font-family:arial,helvetica,sans-serif;font-size:small;color:rgb(0,0,0)">
<br></div>
<div style="font-family:arial,helvetica,sans-serif;font-size:small;color:rgb(0,0,0)">
Are you logging all requests to a central location? When the
problem occurs, it might help taking a closer look at the type of
requests you're receiving.</div>
<div style="font-family:arial,helvetica,sans-serif;font-size:small;color:rgb(0,0,0)">
<br></div>
<div style="font-family:arial,helvetica,sans-serif;font-size:small;color:rgb(0,0,0)">
Depending on how long the mischief lasts, a tcpdump to pcap, then
wireshark might help. Same for an strace on the Puma processes,
similar to what John suggested . Those are low level tools though,
verbose, complex and complete, it's easy to get lost unless you
know what you're looking for.</div>
<div style="font-family:arial,helvetica,sans-serif;font-size:small;color:rgb(0,0,0)">
<br></div>
<div style="font-family:arial,helvetica,sans-serif;font-size:small;color:rgb(0,0,0)">
In summary, CLOSE_WAITs piling up from haproxy (client role) to
Puma (server role) indicates the app not closing connections in
time (or maybe ever) - why? It's a fun one to troubleshoot ;
)</div>
</div>
<div>
<div>
<div class="gmail_extra"><br>
<div class="gmail_quote">On Thu, Jan 14, 2016 at 11:35 PM, Simon
Morley <span dir="ltr"><<a href="mailto:simon@polkaspots.com" target="_blank">simon@polkaspots.com</a>></span> wrote:<br>
<blockquote class="gmail_quote" style="margin:0px 0px 0px 0.8ex;border-left-width:1px;border-left-color:rgb(204,204,204);border-left-style:solid;padding-left:1ex">
<div dir="ltr">Right now, none of the servers have any issues. No
close_waits.
<div><br>
All is well. Seemingly.</div>
<div><br></div>
<div>When it occurs ALL the servers end up going. Sometimes real
fast. That's why I thought we had a db bottleneck. It happens
pretty quickly, randomly, no particular times.</div>
<div><br></div>
<div>We don't ever really get spikes of traffic, there's an even
load inbound throughout.</div>
<div><br></div>
<div>I thought we had someone running a slow loris style attack on
us. So I added some rules to HA Proxy and Cloudflare ain't seen
nofin honest guv.</div>
<div><br></div>
<div>Will find a way to chart it and send a link over.</div>
<div><br></div>
<div>Will see if we're not closing any files - not much of that
going on. There's some manual gzipping happening - we've had that
in place for over a year though - not sure why it'd start playing
up now. Memory usage is high but consistent and doesn't
increase.</div>
<div><span><font color="#888888"><br></font></span></div>
<div><span><font color="#888888">S</font></span></div>
</div>
<div class="gmail_extra"><span><br clear="all"></span>
<div>
<div>
<div dir="ltr">
<div>
<div dir="ltr">
<div><span><br></span></div>
<div><span><br></span></div>
<div>
<div><span>Simon Morley<br>
<span style="font-family:arial;font-size:small"><br>
</span></span></div>
<div><span style="font-family:arial;font-size:small">Big
Chief</span> | PolkaSpots Supafly Wi-Fi
<div>Bigger Chief | Cucumber Tony</div>
<div><br>
<div><span style="font-size:small;font-family:arial">simon@PolkaSpots.com</span><br style="font-size:small;font-family:arial">
<div style="font-size:small;font-family:arial">Linkedin: I'm on it
again and it still sucks</div>
<div style="font-size:small;font-family:arial"><a href="tel:020%207183%201471" value="+442071831471" target="_blank">020
7183 1471</a></div>
</div>
</div>
</div>
</div>
<div style="font-size:small;font-family:arial"><br></div>
<div><font face="arial" size="2">🚀💥</font><br></div>
</div>
</div>
</div>
</div>
</div>
<br>
<div>
<div>
<div class="gmail_quote">On 14 January 2016 at 22:14, Gerhard Lazu
<span dir="ltr"><<a href="mailto:gerhard@lazu.co.uk" target="_blank">gerhard@lazu.co.uk</a>></span> wrote:<br>
<blockquote class="gmail_quote" style="margin:0px 0px 0px 0.8ex;border-left-width:1px;border-left-color:rgb(204,204,204);border-left-style:solid;padding-left:1ex">
<div dir="ltr">
<div style="font-family:arial,helvetica,sans-serif;font-size:small;color:rgb(0,0,0)">
That sounds like a file descriptor leak. Are the CLOSE_WAITs
growing over time?</div>
<div style="font-family:arial,helvetica,sans-serif;font-size:small;color:rgb(0,0,0)">
<br></div>
<div style="font-family:arial,helvetica,sans-serif;font-size:small;color:rgb(0,0,0)">
You're right, New Relic is too high level, this is a layer 4-5
issue.</div>
<div style="font-family:arial,helvetica,sans-serif;font-size:small;color:rgb(0,0,0)">
<br></div>
<div style="font-family:arial,helvetica,sans-serif;font-size:small;color:rgb(0,0,0)">
The simplest thing that can plot some graphs will work. Throw the
dirtiest script together that curls the data out if it comes easy,
it doesn't matter how you get those metrics as long as you have
them.</div>
<div style="font-family:arial,helvetica,sans-serif;font-size:small;color:rgb(0,0,0)">
<br></div>
<div style="font-family:arial,helvetica,sans-serif;font-size:small;color:rgb(0,0,0)">
This is a great blog post opportunity ; )</div>
</div>
<div>
<div>
<div class="gmail_extra"><br>
<div class="gmail_quote">On Thu, Jan 14, 2016 at 8:40 PM, Simon
Morley <span dir="ltr"><<a href="mailto:simon@polkaspots.com" target="_blank">simon@polkaspots.com</a>></span> wrote:<br>
<blockquote class="gmail_quote" style="margin:0px 0px 0px 0.8ex;border-left-width:1px;border-left-color:rgb(204,204,204);border-left-style:solid;padding-left:1ex">
<div dir="auto">
<div>I would ordinarily agree with you about the connection however
they hang around for hours sometimes. </div>
<div><br></div>
<div>The 500 in the hyproxy config was actually left over from a
previous experiment. Realistically I know they won't cope with
that.</div>
<div><br></div>
<div>Using another server was to find any issues with puma. I'm
still going to try unicorn just in case. <br>
<br>
Will up the numbers too - thanks for that suggestion. </div>
<div><br></div>
<div>I'll look at a better monitoring tool too. So far new relic
hasn't helped much. </div>
<div><br></div>
<div>Thanks</div>
<div><br></div>
<div>S</div>
<div><span><br></span>
<div style="font-family:Helvetica;font-size:medium"><span>Simon
Morley</span></div>
<div style="font-family:Helvetica;font-size:medium"><span>Big Chief
| PolkaSpots Supafly Wi-Fi</span></div>
<div style="font-family:Helvetica;font-size:medium">
<span><br></span></div>
<div style="font-family:Helvetica;font-size:medium">
<span><br></span></div>
<div style="font-family:Helvetica;font-size:medium">
<span><br></span></div>
<div style="font-family:Helvetica;font-size:medium">I'm doing it
with Cucumber Tony. Are you?</div>
</div>
<div>
<div>
<div><br>
On 14 Jan 2016, at 20:30, Gerhard Lazu <<a href="mailto:gerhard@lazu.co.uk" target="_blank">gerhard@lazu.co.uk</a>> wrote:<br>
<br></div>
<blockquote type="cite">
<div>
<div dir="ltr">
<div style="font-family:arial,helvetica,sans-serif;font-size:small;color:rgb(0,0,0)">
Hi Simon,</div>
<div style="font-family:arial,helvetica,sans-serif;font-size:small;color:rgb(0,0,0)">
<br></div>
<div style="font-family:arial,helvetica,sans-serif;font-size:small;color:rgb(0,0,0)">
CLOSE_WAIT suggests that Puma is not closing connections fast
enough. The client has asked for the connection to be closed,
but Puma is busy.</div>
<div style="font-family:arial,helvetica,sans-serif;font-size:small;color:rgb(0,0,0)">
<br></div>
<div style="font-family:arial,helvetica,sans-serif;font-size:small;color:rgb(0,0,0)">
Quickest win would be to increase your Puma instances. Unicorn
won't help - or any other Rack web server for the matter.</div>
<div style="font-family:arial,helvetica,sans-serif;font-size:small;color:rgb(0,0,0)">
<br></div>
<div style="font-family:arial,helvetica,sans-serif;font-size:small;color:rgb(0,0,0)">
Based on your numbers, start with 10 Puma instances. Anything more
than 100 connections for a Rails instance is not realistic. I would
personally go with 50, just to be safe. I think I saw 500
conns in your haproxy config, which is way
too optimistic.</div>
<div style="font-family:arial,helvetica,sans-serif;font-size:small;color:rgb(0,0,0)">
<br></div>
<font face="arial, helvetica, sans-serif" size="2">You want metrics
for detailed CPU usage by process, connections open with state
by process, and memory usage, by process. Without these, you're
flying blind. Any suggestions anyone makes without real
metrics - including myself - are just guesses. You'll get
there, but you're making it far too difficult for
yourself.</font></div>
<div dir="ltr"><font face="arial, helvetica, sans-serif" size="2"><br></font></div>
<div dir="ltr"><font face="arial, helvetica, sans-serif" size="2">Let me know how it goes, Gerhard.</font></div>
<div class="gmail_extra"><br>
<div class="gmail_quote">On Thu, Jan 14, 2016 at 3:16 PM, Simon
Morley <span dir="ltr"><<a>simon@polkaspots.com</a>></span>
wrote:<br>
<blockquote class="gmail_quote" style="margin:0px 0px 0px 0.8ex;border-left-width:1px;border-left-color:rgb(204,204,204);border-left-style:solid;padding-left:1ex">
<div dir="ltr">Hello All
<div><br></div>
<div>We've been battling with Puma for a long while now, I'm
looking for some help / love / attention / advice / anything to
prevent further hair loss.</div>
<div><br></div>
<div>We're using it in a reasonably typical Rails 4 application
behind Nginx. <br clear="all">
<div>
<div>
<div dir="ltr">
<div dir="ltr">
<div><br></div>
<div>Over the last 3 months, our requests have gone from 500 rpm to
a little over 1000 depending on the hour. Over this period, we've
been seeing weird CLOSE_WAIT conns appearing in netstat, which
eventually kill the servers.</div>
<div><br></div>
<div>We have 3 Rails servers behind Haproxy running things. Load is
generally even.</div>
<div><br></div>
<div>Running netstat on the servers shows a pile of connections in
the CLOSE_WAIT state with varying recv-q values as so:<br></div>
<div><br></div>
<div>
<div>tcp 2784 0 localhost:58786
localhost:5100
CLOSE_WAIT</div>
<div>tcp 717 0
localhost:35794 localhost:5100
CLOSE_WAIT</div>
<div>tcp 784 0
localhost:55712 localhost:5100
CLOSE_WAIT<br></div>
<div>tcp 0 0
localhost:38639 localhost:5100
CLOSE_WAIT</div>
</div>
<div><br></div>
<div>
<div>That's just a snippet. A wc reveals over 400 of these on each
server.</div>
</div>
<div><br></div>
<div>Puma is running on port 5100 btw. We've tried puma with
multiple threads and a single one - same result. Latest version as
of today.</div>
<div><br></div>
<div>I've checked haproxy and don't see much lingering
around.</div>
<div><br></div>
<div>Only a kill -9 can stop Puma - otherwise, it says something
like 'waiting for requests to finish'</div>
<div><br></div>
<div>I ran GDB to see if I could debug the process however I can't
claim I knew what I was looking at. The processes that seemed
apparent were event machine and mongo.</div>
<div><br></div>
<div>We then ditched EM (we were using the AMQP gem) in favour of
Bunny. That made zero difference.</div>
<div><br></div>
<div>So we upgraded Mongo and Mongoid to the latest versions,
neither of which helped.</div>
<div><br></div>
<div>I thought we might have a bottleneck somewhere - Mongo, ES or
MySQL. But, none of those services seem to have any issues /
latencies.</div>
<div><br></div>
<div>It's also 100% random. Might happen 10 times in an hour, then
not at all for a week.</div>
<div><br></div>
<div>The puma issues on github don't shed much light.</div>
<div><br></div>
<div>I don't really know where to turn at the moment or what to do
next? I was going to resort back to Unicorn but I don't think the
issue is that side and I wanted to fix the problem, not just patch
it up.</div>
<div><br></div>
<div>It's starting to look like a nasty in my code somewhere but I
don't want to go down that route just yet...</div>
<div><br></div>
<div>Sorry for the long email, thanks in advance. Stuff.</div>
<div><br></div>
<div>I hope someone can help!</div>
<div><br></div>
<div>S</div>
<div><br></div>
<div>
<div>Simon Morley<br>
<span style="font-family:arial;font-size:small"><br></span></div>
<div><span style="font-family:arial;font-size:small">Big
Chief</span> | PolkaSpots Supafly Wi-Fi
<div>Bigger Chief | Cucumber Tony</div>
<div><br>
<div><span style="font-size:small;font-family:arial"><a href="mailto:simon@polkaspots.com" target="_blank">simon@PolkaSpots.com</a></span><br style="font-size:small;font-family:arial">
<div style="font-size:small;font-family:arial">Linkedin: I'm on it
again and it still sucks</div>
<div style="font-size:small;font-family:arial"><a href="tel:020%207183%201471" value="+442071831471" target="_blank">020
7183 1471</a></div>
</div>
</div>
</div>
</div>
<div style="font-size:small;font-family:arial"><br></div>
<div><font face="arial" size="2">🚀💥</font><br></div>
</div>
</div>
</div>
</div>
</div>
</div>
<br>
_______________________________________________<br>
Chat mailing list<br>
<a>Chat@lists.lrug.org</a><br>
Archives: <a href="http://lists.lrug.org/pipermail/chat-lrug.org" rel="noreferrer" target="_blank">http://lists.lrug.org/pipermail/chat-lrug.org</a><br>
Manage your subscription: <a href="http://lists.lrug.org/options.cgi/chat-lrug.org" rel="noreferrer" target="_blank">http://lists.lrug.org/options.cgi/chat-lrug.org</a><br>
List info: <a href="http://lists.lrug.org/listinfo.cgi/chat-lrug.org" rel="noreferrer" target="_blank">http://lists.lrug.org/listinfo.cgi/chat-lrug.org</a><br>
<br></blockquote>
</div>
<br></div>
</div>
</blockquote>
<blockquote type="cite">
<div>
<span>_______________________________________________</span><br>
<span>Chat mailing list</span><br>
<span><a href="mailto:Chat@lists.lrug.org" target="_blank">Chat@lists.lrug.org</a></span><br>
<span>Archives: <a href="http://lists.lrug.org/pipermail/chat-lrug.org" target="_blank">http://lists.lrug.org/pipermail/chat-lrug.org</a></span><br>
<span>Manage your subscription: <a href="http://lists.lrug.org/options.cgi/chat-lrug.org" target="_blank">http://lists.lrug.org/options.cgi/chat-lrug.org</a></span><br>
<span>List info: <a href="http://lists.lrug.org/listinfo.cgi/chat-lrug.org" target="_blank">http://lists.lrug.org/listinfo.cgi/chat-lrug.org</a></span><br>
</div>
</blockquote>
</div>
</div>
</div>
<br>
_______________________________________________<br>
Chat mailing list<br>
<a href="mailto:Chat@lists.lrug.org" target="_blank">Chat@lists.lrug.org</a><br>
Archives: <a href="http://lists.lrug.org/pipermail/chat-lrug.org" rel="noreferrer" target="_blank">http://lists.lrug.org/pipermail/chat-lrug.org</a><br>
Manage your subscription: <a href="http://lists.lrug.org/options.cgi/chat-lrug.org" rel="noreferrer" target="_blank">http://lists.lrug.org/options.cgi/chat-lrug.org</a><br>
List info: <a href="http://lists.lrug.org/listinfo.cgi/chat-lrug.org" rel="noreferrer" target="_blank">http://lists.lrug.org/listinfo.cgi/chat-lrug.org</a><br>
<br></blockquote>
</div>
<br></div>
</div>
</div>
<br>
_______________________________________________<br>
Chat mailing list<br>
<a href="mailto:Chat@lists.lrug.org" target="_blank">Chat@lists.lrug.org</a><br>
Archives: <a href="http://lists.lrug.org/pipermail/chat-lrug.org" rel="noreferrer" target="_blank">http://lists.lrug.org/pipermail/chat-lrug.org</a><br>
Manage your subscription: <a href="http://lists.lrug.org/options.cgi/chat-lrug.org" rel="noreferrer" target="_blank">http://lists.lrug.org/options.cgi/chat-lrug.org</a><br>
List info: <a href="http://lists.lrug.org/listinfo.cgi/chat-lrug.org" rel="noreferrer" target="_blank">http://lists.lrug.org/listinfo.cgi/chat-lrug.org</a><br>
<br></blockquote>
</div>
<br></div>
</div>
</div>
<br>
_______________________________________________<br>
Chat mailing list<br>
<a href="mailto:Chat@lists.lrug.org" target="_blank">Chat@lists.lrug.org</a><br>
Archives: <a href="http://lists.lrug.org/pipermail/chat-lrug.org" rel="noreferrer" target="_blank">http://lists.lrug.org/pipermail/chat-lrug.org</a><br>
Manage your subscription: <a href="http://lists.lrug.org/options.cgi/chat-lrug.org" rel="noreferrer" target="_blank">http://lists.lrug.org/options.cgi/chat-lrug.org</a><br>
List info: <a href="http://lists.lrug.org/listinfo.cgi/chat-lrug.org" rel="noreferrer" target="_blank">http://lists.lrug.org/listinfo.cgi/chat-lrug.org</a><br>
<br></blockquote>
</div>
<br></div>
</div>
</div>
<br>
_______________________________________________<br>
Chat mailing list<br>
<a href="mailto:Chat@lists.lrug.org" target="_blank">Chat@lists.lrug.org</a><br>
Archives: <a href="http://lists.lrug.org/pipermail/chat-lrug.org" rel="noreferrer" target="_blank">http://lists.lrug.org/pipermail/chat-lrug.org</a><br>
Manage your subscription: <a href="http://lists.lrug.org/options.cgi/chat-lrug.org" rel="noreferrer" target="_blank">http://lists.lrug.org/options.cgi/chat-lrug.org</a><br>
List info: <a href="http://lists.lrug.org/listinfo.cgi/chat-lrug.org" rel="noreferrer" target="_blank">http://lists.lrug.org/listinfo.cgi/chat-lrug.org</a><br>
<br></blockquote>
</div>
<br></div>
</div>
</div>
<br>
_______________________________________________<br>
Chat mailing list<br>
<a href="mailto:Chat@lists.lrug.org" target="_blank">Chat@lists.lrug.org</a><br>
Archives: <a href="http://lists.lrug.org/pipermail/chat-lrug.org" rel="noreferrer" target="_blank">http://lists.lrug.org/pipermail/chat-lrug.org</a><br>
Manage your subscription: <a href="http://lists.lrug.org/options.cgi/chat-lrug.org" rel="noreferrer" target="_blank">http://lists.lrug.org/options.cgi/chat-lrug.org</a><br>
List info: <a href="http://lists.lrug.org/listinfo.cgi/chat-lrug.org" rel="noreferrer" target="_blank">http://lists.lrug.org/listinfo.cgi/chat-lrug.org</a><br>
<br></blockquote>
</div>
<br></div>
</div>
</div>
</div>
<br>
_______________________________________________<br>
Chat mailing list<br>
<a href="mailto:Chat@lists.lrug.org" target="_blank">Chat@lists.lrug.org</a><br>
Archives: <a href="http://lists.lrug.org/pipermail/chat-lrug.org" rel="noreferrer" target="_blank">http://lists.lrug.org/pipermail/chat-lrug.org</a><br>
Manage your subscription: <a href="http://lists.lrug.org/options.cgi/chat-lrug.org" rel="noreferrer" target="_blank">http://lists.lrug.org/options.cgi/chat-lrug.org</a><br>
List info: <a href="http://lists.lrug.org/listinfo.cgi/chat-lrug.org" rel="noreferrer" target="_blank">http://lists.lrug.org/listinfo.cgi/chat-lrug.org</a><br>
<br></blockquote>
</div>
<br></div>
</div>
</div>
_______________________________________________
<br>Chat mailing list
<br><a href="mailto:Chat@lists.lrug.org" target="_blank">Chat@lists.lrug.org</a>
<br>Archives: <a href="http://lists.lrug.org/pipermail/chat-lrug.org" target="_blank">http://lists.lrug.org/pipermail/chat-lrug.org</a>
<br>Manage your subscription: <a href="http://lists.lrug.org/options.cgi/chat-lrug.org" target="_blank">http://lists.lrug.org/options.cgi/chat-lrug.org</a>
<br>List info: <a href="http://lists.lrug.org/listinfo.cgi/chat-lrug.org" target="_blank">http://lists.lrug.org/listinfo.cgi/chat-lrug.org</a>
<br></div></div></span></blockquote></div></div></div></div></div></blockquote></div></div></div><br></div>
<br>_______________________________________________<br>
Chat mailing list<br>
<a href="mailto:Chat@lists.lrug.org" target="_blank">Chat@lists.lrug.org</a><br>
Archives: <a href="http://lists.lrug.org/pipermail/chat-lrug.org" rel="noreferrer" target="_blank">http://lists.lrug.org/pipermail/chat-lrug.org</a><br>
Manage your subscription: <a href="http://lists.lrug.org/options.cgi/chat-lrug.org" rel="noreferrer" target="_blank">http://lists.lrug.org/options.cgi/chat-lrug.org</a><br>
List info: <a href="http://lists.lrug.org/listinfo.cgi/chat-lrug.org" rel="noreferrer" target="_blank">http://lists.lrug.org/listinfo.cgi/chat-lrug.org</a><br>
<br></blockquote></div><br><br clear="all"><div><br></div></div></div><span><font color="#888888">-- <br><div><div dir="ltr">Riccardo Tacconi<br><br><a href="http://github.com/rtacconi" target="_blank">http://github.com/rtacconi</a><br><a href="http://twitter.com/rtacconi" target="_blank">http://twitter.com/rtacconi</a></div></div>
</font></span></div>
<br>_______________________________________________<br>
Chat mailing list<br>
<a href="mailto:Chat@lists.lrug.org" target="_blank">Chat@lists.lrug.org</a><br>
Archives: <a href="http://lists.lrug.org/pipermail/chat-lrug.org" rel="noreferrer" target="_blank">http://lists.lrug.org/pipermail/chat-lrug.org</a><br>
Manage your subscription: <a href="http://lists.lrug.org/options.cgi/chat-lrug.org" rel="noreferrer" target="_blank">http://lists.lrug.org/options.cgi/chat-lrug.org</a><br>
List info: <a href="http://lists.lrug.org/listinfo.cgi/chat-lrug.org" rel="noreferrer" target="_blank">http://lists.lrug.org/listinfo.cgi/chat-lrug.org</a><br>
<br></blockquote></div><br></div></div></div>
<br>_______________________________________________<br>
Chat mailing list<br>
<a href="mailto:Chat@lists.lrug.org" target="_blank">Chat@lists.lrug.org</a><br>
Archives: <a href="http://lists.lrug.org/pipermail/chat-lrug.org" rel="noreferrer" target="_blank">http://lists.lrug.org/pipermail/chat-lrug.org</a><br>
Manage your subscription: <a href="http://lists.lrug.org/options.cgi/chat-lrug.org" rel="noreferrer" target="_blank">http://lists.lrug.org/options.cgi/chat-lrug.org</a><br>
List info: <a href="http://lists.lrug.org/listinfo.cgi/chat-lrug.org" rel="noreferrer" target="_blank">http://lists.lrug.org/listinfo.cgi/chat-lrug.org</a><br>
<br></blockquote></div><br><br clear="all"><div><br></div>-- <br><div><div dir="ltr">Riccardo Tacconi<br><br><a href="http://github.com/rtacconi" target="_blank">http://github.com/rtacconi</a><br><a href="http://twitter.com/rtacconi" target="_blank">http://twitter.com/rtacconi</a></div></div>
</div></div></div></div></div>
<br>_______________________________________________<br>
Chat mailing list<br>
<a href="mailto:Chat@lists.lrug.org" target="_blank">Chat@lists.lrug.org</a><br>
Archives: <a href="http://lists.lrug.org/pipermail/chat-lrug.org" rel="noreferrer" target="_blank">http://lists.lrug.org/pipermail/chat-lrug.org</a><br>
Manage your subscription: <a href="http://lists.lrug.org/options.cgi/chat-lrug.org" rel="noreferrer" target="_blank">http://lists.lrug.org/options.cgi/chat-lrug.org</a><br>
List info: <a href="http://lists.lrug.org/listinfo.cgi/chat-lrug.org" rel="noreferrer" target="_blank">http://lists.lrug.org/listinfo.cgi/chat-lrug.org</a><br>
<br></blockquote></div><br></div></div></div>
<br>_______________________________________________<br>
Chat mailing list<br>
<a href="mailto:Chat@lists.lrug.org" target="_blank">Chat@lists.lrug.org</a><br>
Archives: <a href="http://lists.lrug.org/pipermail/chat-lrug.org" rel="noreferrer" target="_blank">http://lists.lrug.org/pipermail/chat-lrug.org</a><br>
Manage your subscription: <a href="http://lists.lrug.org/options.cgi/chat-lrug.org" rel="noreferrer" target="_blank">http://lists.lrug.org/options.cgi/chat-lrug.org</a><br>
List info: <a href="http://lists.lrug.org/listinfo.cgi/chat-lrug.org" rel="noreferrer" target="_blank">http://lists.lrug.org/listinfo.cgi/chat-lrug.org</a><br>
<br></blockquote></div><br><br clear="all"><div><br></div>-- <br><div><div dir="ltr">Riccardo Tacconi<br><br><a href="http://github.com/rtacconi" target="_blank">http://github.com/rtacconi</a><br><a href="http://twitter.com/rtacconi" target="_blank">http://twitter.com/rtacconi</a></div></div>
</div>
</div></div><br>_______________________________________________<br>
Chat mailing list<br>
<a href="mailto:Chat@lists.lrug.org">Chat@lists.lrug.org</a><br>
Archives: <a href="http://lists.lrug.org/pipermail/chat-lrug.org" rel="noreferrer" target="_blank">http://lists.lrug.org/pipermail/chat-lrug.org</a><br>
Manage your subscription: <a href="http://lists.lrug.org/options.cgi/chat-lrug.org" rel="noreferrer" target="_blank">http://lists.lrug.org/options.cgi/chat-lrug.org</a><br>
List info: <a href="http://lists.lrug.org/listinfo.cgi/chat-lrug.org" rel="noreferrer" target="_blank">http://lists.lrug.org/listinfo.cgi/chat-lrug.org</a><br>
<br></blockquote></div><br></div>