Large changes in Pagespeed Insights scores and warnings

[lldt]

We've been getting varying Pagespeed Insights results (from 20 - 75), but more importantly, warnings come and go. For example. images (and other resources) which were showing as optimized and not warning on one run of the test, 5 mins later, without changing anything on the site, can show up as not optimized.  I'm just wondering if there is some server caching that I'm not aware of. There are no related errors in the apache logs. The setup is a Bitnami Wordpress stack on EC2 with Lets Encrypt SSL, Ubuntu 14.04, Apache 2.4.23 and mod_pagespeed 1.9.32.14-0 (can't update this without building from source).

Pagespeed.conf:

   ModPagespeedRewriteLevel CoreFilters

   ModPagespeedEnableFilters convert_png_to_jpeg,convert_jpeg_to_webp

   ModPagespeedEnableFilters collapse_whitespace,remove_comments,combine_css

   ModPagespeedEnableCachePurge on

   ModPagespeedMapOriginDomain "http://somedomain.com" "https://somedomain.com

Had some trouble with getting SSL to work, so using ModPagespeedMapOriginDomain for now. I think I'll turn off ModPagespeedEnableCachePurge, but that's not the cause. Note there is a CNAME of www.somedomain.com to somedomain.com

Any ideas?

[lldt]

I've since added a few more items to the conf:

ModPagespeedImplicitCacheTtlMs 50000000

ModPagespeedFetchWithGzip on

SetOutputFilter DEFLATE

ModPagespeedEnableFilters extend_cache 

I *think* the issue is partially the timeout (ModPagespeedImplicitCacheTtlMs) and the fact that viewing the site on desktop generates the optimized webp version, but that image is only displayed on desktop. You have to access the site on mobile before it STARTS to generate one for mobile (then you need to reload the page before it starts using it). So if no mobile users visit the site for a while, all the optimized resources expire and mobile PageSpeed Insights will report a bad score.

[Joshua Marantz]

Correct.  mod_pagespeed works best on sites with traffic.  One workaround is to make a cron-job that crawls the site every minute with various user-agents, to keep the server-side cache fresh.

--
You received this message because you are subscribed to the Google Groups "mod-pagespeed-discuss" group.
To unsubscribe from this group and stop receiving emails from it, send an email to mod-pagespeed-discuss+unsub...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/mod-pagespeed-discuss/1ae9cbd1-287c-4f24-8ec2-0bbf49e25bd7%40googlegroups.com.

For more options, visit https://groups.google.com/d/optout.

[lldt]

Ok thanks. I've already increased the ModPagespeedImplicitCacheTtlMs setting. Hopefully that will solve this. Creating a cron job to crawl the site with various user agents sounds like a last resort thing, so hopefully it won't come down to that.

According to documentation, ModPagespeedImplicitCacheTtlMs is supposed to be based on the expires value of the resource and only falls back to its default of 5 mins if no expiry is found, but after turning on mod_expires and putting in the correct expire statements and in the correct place, nothing seems to happen. I've read about other people experiencing the same thing with this Bitnami stack. Really strange.