<div dir="ltr">In my experience, it's usually a few naughty requests that gobble up a lot of memory. We found it useful to record process vsize and rss at the beginning and end of the request cycle to generate warnings like this:<div><br></div><div>Warning: PID 28710 grew rss by 20% from 42 MiB to 51 MiB serving GET /api/search?q=foo</div><div><br></div><div>One of the common mistakes in our app was inflating all row objects in a result set at the same time when it wasn't needed. We prefetch a lot so row objects and their related objects get big. In those cases, this:</div><div><br></div><div><div><span class="" style="white-space:pre"> </span>foreach my $row ($rs->all)</div><div><span class="" style="white-space:pre"> </span>{</div><div><span class="" style="white-space:pre"> </span>push @results, $row->to_json,</div><div><span class="" style="white-space:pre"> </span>}</div><div><br></div><div>Will use a lot more memory than this:</div><div><br></div><div><span class="" style="white-space:pre"> </span>while (my $row = $rs->next)</div><div><span class="" style="white-space:pre"> </span>{</div><div><span class="" style="white-space:pre"> </span>push @results, $row->to_json,</div><div><span class="" style="white-space:pre"> </span>}</div></div><div><br></div><div>DBIx::Class's HashRefInflator is often touted for its speed but it's also a lot lighter on memory...</div><div><br></div><div>Another thing you could consider is using Plack to run all your Perl apps in the same server/process. Memory for all common dependencies would be shared among the apps in that case which might get you a decent gain...</div><div><br></div><div>There's also a bunch of Plack::MiddleWare's on CPAN to help with debugging/profiling of apps...</div><div><br></div><div>/L<br></div><div><br></div><div><br><div class="gmail_extra"><br><div class="gmail_quote">On Thu, Jun 11, 2015 at 3:03 PM, Denny <span dir="ltr"><<a href="mailto:2015@denny.me" target="_blank">2015@denny.me</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>I run a server with 18 separate installs of ShinyCMS running, with at least 3 and sometimes 20 processes per install. 122 perl-fcgi processes in total. They look a bit bigger than most of your processes. I've got 8GB of RAM, and I'm pushing up against the edges of it a bit, but I don't usually have any problems.<br>
<br>
Hope that's useful info!<br>
<br>
<br><br><div class="gmail_quote"><div><div>On 11 June 2015 13:31:16 BST, Jesse Sheidlower <<a href="mailto:jester@panix.com" target="_blank">jester@panix.com</a>> wrote:</div></div><blockquote class="gmail_quote" style="margin:0pt 0pt 0pt 0.8ex;border-left-width:1px;border-left-style:solid;border-left-color:rgb(204,204,204);padding-left:1ex">
<pre><div><div><br>Sysadmin-fu needed: I have a personal server in the cloud that has 1 GB<br>of RAM. It's running Debian, and the usual server basics--MySQL, nginx,<br>etc. No X, of course.<br><br>There are three Catalyst apps running, all of them fairly small, using<br>nginx and FCGI (I run the MyApp_<a href="http://fastcgi.pl" target="_blank">fastcgi.pl</a> script to a socket, that<br>nginx then responds to). But they're taking up all the memory on the<br>system, and I can't run another Cat app (and sometimes even system<br>commands hang for lack of memory). I tried reducing --nproc to 1 for two<br>of the three apps that I'm the only one using (the other has --nproc 3).<br>The output of "top" with "M" (sorting by RES) looks like (I removed a<br>few columns for spacing):<br><br> PID USER VIRT RES SHR S %MEM TIME+ COMMAND <br><br>23766 www-data 440m 230m 2584 S 23.0 0:13.40 perl-fcgi <br>16958 mysql 428m 188m 0 S 18.8 118:14.98
mysqld <br>25992 www-data 375m 158m 2608 S 15.9 0:06.80 perl-fcgi <br>26884 www-data 238m 80m 2616 S 8.0 0:01.61 perl-fcgi <br> 2513 www-data 282m 76m 0 S 7.7 0:06.01 perl-fcgi <br> 2512 www-data 182m 66m 0 S 6.6 0:00.00 perl-fcgi-pm [Library]<br> 1825 www-data 172m 61m 0 S 6.1 0:03.44 perl <br> 2487 www-data 163m 46m 60 S 4.6 0:00.18 perl-fcgi-pm [CatSF]<br> 1692 root 119m 5128 140 S 0.5 0:56.32 rsyslogd <br>19998 www-data 79452 3296 1136 S 0.3 6:10.61 nginx <br> 2265 www-data 79768 3148 504 S 0.3 15:57.81 nginx <br>[...]<br><br>Is this an expected or appropriate amount of memory usage? I appreciate<br>that 1GB isn't all that much memory nowadays, but this is a very lightly<br>loaded machine, and I don't want to pay a lot more for a larger cloud<br>server to handle this if it's not necessary.<br><br>Thanks.<br><br><hr><br></div></div><span>List: <a href="mailto:Catalyst@lists.scsys.co.uk" target="_blank">Catalyst@lists.scsys.co.uk</a><br>Listinfo: <a href="http://lists.scsys.co.uk/cgi-bin/mailman/listinfo/catalyst" target="_blank">http://lists.scsys.co.uk/cgi-bin/mailman/listinfo/catalyst</a><br>Searchable archive: <a href="http://www.mail-archive.com/catalyst@lists.scsys.co.uk" target="_blank">http://www.mail-archive.com/catalyst@lists.scsys.co.uk</a>/<br>Dev site: <a href="http://dev.catalyst.perl.org" target="_blank">http://dev.catalyst.perl.org</a>/<br></span></pre></blockquote></div><span><font color="#888888"><br>
-- <br>
Sent from my phone. Please excuse typos, terseness, and top-posting.</font></span></div><br>_______________________________________________<br>
List: <a href="mailto:Catalyst@lists.scsys.co.uk" target="_blank">Catalyst@lists.scsys.co.uk</a><br>
Listinfo: <a href="http://lists.scsys.co.uk/cgi-bin/mailman/listinfo/catalyst" rel="noreferrer" target="_blank">http://lists.scsys.co.uk/cgi-bin/mailman/listinfo/catalyst</a><br>
Searchable archive: <a href="http://www.mail-archive.com/catalyst@lists.scsys.co.uk/" rel="noreferrer" target="_blank">http://www.mail-archive.com/catalyst@lists.scsys.co.uk/</a><br>
Dev site: <a href="http://dev.catalyst.perl.org/" rel="noreferrer" target="_blank">http://dev.catalyst.perl.org/</a><br>
<br></blockquote></div><br></div></div></div>