portal entry

select a category, or use search below
(searches all categories and all time range)
Title:

Anyone else experiencing performance issues over CF2018/Apache

| View in Portal
March 07, 2019 03:06:09 PM GMT
3 Comments
<p>Hi community! Our team has been using ColdFusion at least since CF7MX and probably even prior to that. We’re currently migrating from CF10 to CF2018. Our configuration is simple: Linux, Apache, ColdFusion and Oracle (LACFO? No? Well, I tried). Over the years this configuration has served us well, and even though we had to do a bit of research and configuration when migrating to CF10 we never had really bad performance. With the new servers we got however we’re seeing […]</p>
<p>The post <a rel="nofollow" href="https://coldfusion.adobe.com/2019/03/anyone-else-experiencing-performance-issues-cf2018-apache/">Anyone else experiencing performance issues over CF2018/Apache</a> appeared first on <a rel="nofollow" href="https://coldfusion.adobe.com">ColdFusion</a>.</p>
Labels: CF2018 Updates, ColdFusion 2018, Discussion, 2018, cf2018 updates, coldfusion 2018, discussion, performance, server

Comments:

Miguel, there is no inherent reason cf 2018 would be slower, no. First can you confirm if it was slow with no updates, and/or update 1 or 2 or 3? More important, you refer to these as "new servers", so do you really mean that cf2018 is running on a new box compared to cf10? If so, then first note that the change in performance could be related to that. What else changed? OS version, too? Apache version? Of course the jvm version changed  as cf2018 runs on Java 10--or 11 as of  the new installers in Feb. Have you confirmed that any tweaks you'd made to those on the old are carried forward to the new? Or perhaps some tweaks you made to make things non-standard on the old are no longer appropriate to the new (especially perhaps jvm tweaks). Have you indeed made sure to raise cf's max heap size z as the default of 1g is too small for most production needs, in my experience. Finally, do you have any monitoring on CF? Whether the cf2018 PMT, FusionReactor, jvm tools, or others, you can't really diagnose what you can't measure. And if those clues aren't enough and you need this solved, I'll add that I help people with such problems every day, in my consulting (carehart.org/consulting), remotely, safely, quickly, and with satisfaction guaranteed or you won't pay for time you don't value. Happy to help here, of course, while feasible. And others will likely chime in.
Comment by Charlie Arehart
1885 | March 08, 2019 02:36:20 PM GMT
Hi Charlie, We updated up to patch 2 fairly quickly so prior to that I can't say. We're now on update 3; still the same story. Yes, we have entirely new servers running CF2018. These are significantly beefier than the ones we have running CF10 when it comes to RAM and CPU. We switched from Debian (Wheezy) to RHEL7 and from Apache 2.2 to Apache 2.4. I increased the max heap size to 8GB and set the min to 1GB. I've been trying different JVM settings to no avail. Right now we don't have a monitoring tool, but we've been using Chrome's Page Load Time extension to track and compare across the board. We're aware this is not very robust. What we've noticed is that simple Requests times (as in the time between server getting the request and "first byte") are significantly lower on the new systems, it is only when we do more complex operations that the performance becomes worse than the old configuration, the bigger difference being the Requests time. Response times (as in time from "first byte" to the moment the browser starts rendering the DOM) on the other hand seem to be better on the old ones but that might tweaks on the Apache configuration.
Comment by Miguel Arvelo
1899 | March 08, 2019 04:34:58 PM GMT
You will need monitoring within cf to know of that slow response time is in cf (as you may be presuming) or perhaps in/due to the web connector (between apache and CF), or perhaps even in apache itself (it can happen). I could help you know (implementing the fr free trial, with a CF restart, then determining if the slowness is in CF) in as little as 15 mins together, if interested.
Comment by Charlie Arehart
1903 | March 09, 2019 01:17:19 AM GMT