Created attachment 13970 [details]
Firefox inspector timing for page load

Confirmed. Checking with cache disabled in firefox/inspector shows it's
waiting 10 seconds after the tls handshake and then taking 3ms to actually
transfer the data.

CC: (none) => davidwhodgins

Adding sysadmin team to cc list as it looks like a server issue.

CC: (none) => sysadmin-bugs

looks that related to div that get blog post https://gitweb.mageia.org/web/www/tree/lib/news.php
same issue now on https://www.mageia.org/en as I added them back (previously removed as it was telling about RC1 until this morning)

Same it cannot get that it needs

removing again from main page.

Strange issue, as it work fine on my test server.

www.mageia.org is in the LOST-OASIS data center while the blog comes from the
SCALEWAY data center.

Anyway, while removing the blog from https://www.mageia.org/en/community/
leaves a large blank area, the loading time has increased to 25.60 s.

Something is still very wrong.

does it work better now ?

No. Firefox inspector with cache disabled shows it's currently 25.60 seconds.

(In reply to Thomas Backlund from comment #5)
> does it work better now ?

It works a few worst :(

I suggest move the planet parse code to a php script, store the result in file, include_once the file in the right part, cron the php script to be run 2 or 3 times at day (i think we not generate a load of post that need a high number)

its suppose to create a caches files in _nav/var/tmp/cache
http://gitweb.mageia.org/web/www/tree/lib/news.php#n71

that we can delete with 
https://wiki.mageia.org/en/Managing_the_website#Options_for_deleting_feed_and_navigation_cache

http://gitweb.mageia.org/web/www/tree/test/lang.php#n13

but on my server I don't have necessary these files present or recreated, but even if they are missing or not there is no delay on the webpage.

So I don't understand right now

If nothing has changed on server side I will disable it and tests later

Ok, the curl extension is not enable so it use fsockopen
https://www.mageia.org/test/sp_test.php

But there is an error:
fsockopen error: Connection timed out

https://mga.hiebel.eu/test/news.php

On my server, w/ and w/o the curl extension, it works

Sorry last link should be https://mageia.org/test/news.php

(In reply to Manuel Hiebel from comment #10)
> Sorry last link should be https://mageia.org/test/news.php

Now the Community page load time is good, the test page delay 00:11:55 to load

Loading https://www.mageia.org/test/news.php takes 12.80 s, displaying the
graphical header and then a text box starting with ...
NULL
int(1575832)
int(1670600)
SimplePie Object
(
    [data] => Array
        (
        )

    [error] => fsockopen error: Connection timed out
    [status_code] => 
    [sanitize] => SimplePie_Sanitize Object
<snip rest>

The community page is now loading in 1.92 s, with the "Planet | Blog | Forums"
links in the dialog on the left side.

this morning it works fine, no "Connection timed out", did a sysadmin made some changes ?

(In reply to Manuel Hiebel from comment #13)
> this morning it works fine, no "Connection timed out", did a sysadmin made
> some changes ?

The planet part has been restored and load time is good, so i close the bug
Thank to all for your attention even when i not was sure if exist a issue

Status: NEW => RESOLVED
Resolution: (none) => FIXED

I think you should deep check the server were is the planet, again is slowing the load time of main and community page

GET
https://www.mageia.org/es/community/
[HTTP/1.1 200 OK 21022ms]

Resolution: FIXED => (none)
Status: RESOLVED => REOPENED

It looks as if the ipv6 connection between LOST-OASIS and SCALEWAY is failing
so it's timing out and then falling back to ipv4.

yeah, a puppet bug caused a failed update for neru ipv6, does it work better now ?

(I dont have ipv6 so I cant test myself)

Checked my router and it has ipv6 working again, so I re-enabled it in my
install. It wasn't working, so I had disabled ipv6 in sysctl and firefox
when I was testing it for the prior comments.

https://www.mageia.org/en/ took 15.26s
https://www.mageia.org/en/community/ took 12.80s
https://blog.mageia.org/en/ took 5.12s

While the blog on neru has improved noticeably, the main site on sucuk is
still slower than what it used to be.

Is ipv6 working on sucuk?

Ignore comment 18.
[dave@x3 ~]$ ping -4 -c 1 sucuk.mageia.org|grep from
64 bytes from sucuk.mageia.org (212.85.158.151): icmp_seq=1 ttl=41 time=124 ms
[dave@x3 ~]$ ping -6 -c 1 sucuk.mageia.org|grep from
64 bytes from sucuk.mageia.org (2a02:2178:2:7::7): icmp_seq=1 ttl=46 time=120 ms
[dave@x3 ~]$ ping -4 -c 1 neru.mageia.org|grep from
64 bytes from neru.mageia.org (163.172.148.228): icmp_seq=1 ttl=47 time=105 ms
[dave@x3 ~]$ ping -6 -c 1 neru.mageia.org|grep from
64 bytes from 2001:bc8:634:240e::1 (2001:bc8:634:240e::1): icmp_seq=1 ttl=45 time=108 ms

So it's working ok on both. I'm not sure why sucuk is slow loading the page,
but it's probably just the web page itself rather then a networking issue.

GET
https://www.mageia.org/es/community/
[HTTP/1.1 200 OK 20837ms]

hum, reading this bugreport again I see simplepie wants curl, so php-curl is now installed.

does it work better for you ?

(In reply to Thomas Backlund from comment #21)
> hum, reading this bugreport again I see simplepie wants curl, so php-curl is
> now installed.
> 
> does it work better for you ?

yes

GET
https://www.mageia.org/es/community/
[HTTP/1.1 200 OK 840ms]

Much better.

In firefox I right click on the page and select inspect. After clicking on
the network header and then the reload for the page (with cache disabled
as part of inspect), it's now 1.60s

Closing as fixed.

Resolution: (none) => FIXED
Status: REOPENED => RESOLVED