openSUSE forum website load very slowly. it takes long time to load completely. when i compare it with other forums, i see that it is very slow. please solve this issue.
On Sat, 01 Sep 2012 15:06:02 +0000, ilAli wrote:
> openSUSE forum website load very slowly. it takes long time to load
> completely. when i compare it with other forums, i see that it is very
> slow. please solve this issue.
If you read through other threads in this forum, you’ll note that we’re
aware of intermittent issues, but at the time of your report, we were not
experiencing any known issues that I’m aware of. Perhaps the issue is
somewhere between you and the server - and in a case like that, there’s
not a lot we can do about it because we don’t control the entire
Internet. 
“very slowly” isn’t easily quantifiable - we need specific data, not just
reports that things are slow. Dates, times, and load times when you see
the problem are a good place to start.
Depending on which browser you’re using, you can also easily grab the
response headers when the page loads - the ‘Via’ header is the single
most useful piece of information you can provide us with.
Jim
–
Jim Henderson
openSUSE Forums Administrator
Forum Use Terms & Conditions at http://tinyurl.com/openSUSE-T-C
i live in iran. i use an adsl 128 kb internet connection. it offten takes about 20-40 seconds to load pages with firefox 15.
with same specifications, i can open other websites and forums much more faster.
i don’t know what is “via” and how get that?
On Sat, 01 Sep 2012 20:16:02 +0000, ilAli wrote:
> i live in iran. i use an adsl 128 kb internet connection. it offten
> takes about 20-40 seconds to load pages with firefox 15.
> with same specifications, i can open other websites and forums much more
> faster.
> i don’t know what is “via” and how get that?
What browser are you using?
Jim
–
Jim Henderson
openSUSE Forums Administrator
Forum Use Terms & Conditions at http://tinyurl.com/openSUSE-T-C
Intermittent here (UK) with response times ranging from about 2 to 30 seconds. Times between about 4 and 6:30 GMT. All other sites I’ve tried responding quickly.
Using Firefox 15 on openSuse 12.1. Web developer extension gives the following headers:
Date: Sun, 02 Sep 2012 06:14:51 GMT Server: Apache/2.2.3 (Linux/SUSE) X-Powered-By: PHP/5.2.5 Cache-Control: private Pragma: private X-UA-Compatible: IE=Edge Content-Type: text/html; charset=UTF-8 Set-Cookie: bb_lastvisit=1346566491; expires=Mon, 02-Sep-2013 06:14:51 GMT; path=/ bb_lastactivity=0; expires=Mon, 02-Sep-2013 06:14:51 GMT; path=/ skimlinks_enabled=1 vbseo_loggedin=deleted; expires=Sat, 03-Sep-2011 06:14:51 GMT; path=/ ZNPCQ003-32323600=1fe420a3; Path=/; Domain=.opensuse.org Via: 1.1 forums.opensuse.org (Access Gateway-ag-BFC678DFBE6884C5-9494665) Keep-Alive: timeout=2, max=100 Connection: Keep-Alive Transfer-Encoding: chunked
I have recently taken over the servers running the forums software. I agree that they are in desperate need for some optimization. Take a look at this (run from Ireland):
WebPagetest Test Details - Dublin : forums.opensuse.org - 09/02/12 13:43:57
I can see why this might be a lot more painful in EMEA than it would be in AMERICAS. Here are a few optimizations that I see, in order of how much I think they can help:
- Sprites
- CDN
- Compression
- Time to first byte (i.e. database and server optimization)
- Caching settings
I’ll start on #3 and #5 today, as those are easy ones to change, and I’ll looking at what I can do for #4 this week. Also, I’m going to propose that we leverage our CDN for forums, but that may be a bit down the road. I’m also going to talk to those who are responsible for the theme and see what we can do about sprites.
FYI, I took care of items #3 and #5. I also made inroads on #4 by adjusting some database settings. My testing from the above site are showing about 30% improvement in page loading time from EMEA and APAC. If anyone can confirm if things are looking noticeably better, I would appreciate it.
The time to get the first resource (forum.php) has dropped by about 15%, but it’s much still much higher than it should be. Since it’s all application/database thinking time, it’s going to be pretty much the same no matter where you are. I’m hoping to get that portion of loading time cut by 1 to 1.5 seconds, which will put it where it should be.
The use of sprites and combining javascript and CSS is going to make the biggest difference for anyone who is not in the AMERICAS region, so I’m going to propose it to those guys this week.
On 09/03/2012 02:16 AM, MatthewEhle wrote:
> If anyone can confirm if things are looking noticeably better, I would
> appreciate it.
openSUSE 11.4, Chrome, wired DSL
accessing from Denmark, 0727 UTC Monday 3 Sep 12
8.74 seconds…which is BLAZINGLY fast, compared to the last six months
or more…
- and, then once there going to the login page:
http://forums.opensuse.org/login/icslogin.php?destination=/member.php?action=getinfo&u=41199
6.61 seconds (super fast! comparatively speaking in relation to the “usual”)
- filled in ID/Pass and pressed enter
10.31 seconds to new page, showing me logged in (extremely FAST!!
[comparatively])
so, to me it looks like you are a genius in being able to fix a long
term problem in just a few hours, so
-=T H A N K Y O U !!!=-
however a rainy day thing to look at when you might have time and interest:
after login the page landed on is either incorrect or
confused…BECAUSE, i was looking at the user page of when i clicked
the login, and i would expect once logged in to be returned to that
page, but instead i land on
http://forums.opensuse.org/member.php?action=getinfo
which wrongly says about MatthewEhle:
“Message
This user has not registered and therefore does not have a profile to view.”
afaik the intermittent wrong-page-landing problem is years old…
–
dd
–
dd http://tinyurl.com/DD-Caveat http://tinyurl.com/DD-Hardware
http://tinyurl.com/DD-Software
What does DistroWatch write about YOU?: http://tinyurl.com/SUSEonDW
On 2012-09-03 09:44, dd@home.dk wrote:
> On 09/03/2012 02:16 AM, MatthewEhle wrote:
>> If anyone can confirm if things are looking noticeably better, I would
>> appreciate it.
>
> openSUSE 11.4, Chrome, wired DSL
>
> accessing from Denmark, 0727 UTC Monday 3 Sep 12
>
> 1. going to: http://forums.opensuse.org/member.php?userid=41199
>
> 8.74 seconds…which is BLAZINGLY fast, compared to the last six months or more…
30 seconds and counting, here (FF). Blank page still as I write this (nntp). I stop the page, click
reload. Same thing.
And I logged in and out a few minutes ago, to see a private message…
Page does not load at all.
I try the main forum page instead, loads in seconds. I click login… nothing changes, no prompt to
login. Waiting for login.novell.com, I read on the status bar at the bottom.
I don’t see any progress 
(still waiting on the login page to load, after a minute or more)
And now my connection died, and refuses to restart :-/
–
Cheers / Saludos,
Carlos E. R.
(from 11.4 x86_64 “Celadon” (Minas Tirith))
On 2012-09-03 13:51, Carlos E. R. wrote:
> And now my connection died, and refuses to restart :-/
I had to reboot to restore my connection. I can connect, as you see my message, but login to the
forum takes more than half a minute to load. Now I get the prompt. And now I login, in a reasonable
time. Firefox does not prompt me for my master password to store the login data: which means that
the login page is not recognized by firefox as a login page and does not stores data. Same behaviour
if I try login to bugzilla, and the password was stored and prompt for my master password worked two
days ago.
Something is broken.
–
Cheers / Saludos,
Carlos E. R.
(from 11.4 x86_64 “Celadon” (Minas Tirith))
it seems the problem is solved for me (or at least it is less than before). i hope this improvement be permanent.
thank you.
I’m almost positive that this is the issue that Jim and I discussed before, where there is some issue going on with the TCP handshake between the client and server. It’s relatively uncommon, but it’s a PITA when it happens. I’m still trying to figure out the root cause and if there are any kernel/network settings that I can adjust. If you want to confirm it, you can get a packet capture when it happens and look for a whole bunch of duplicate ACKs coming from the server right then. I could also take a look at any packet captures, but remember that I have the TLS private keys, so make sure to not send anything you don’t want me to see 
I’ll forward your comment about the login page to the team that designed it.
Thanks to you and ilAli for getting back to me on that.
As for the login issue, I guess I wasn’t aware of it, as I have always just logged in from the main page. I’ll take a look at it this week, as it may not be too difficult to solve. I noticed that you started a new thread about the logout issue, so I don’t think it’s a secret with you that the single login/logout really needs some work. I can dabble in PHP, but this one is way beyond my skill level. Kim and I have been talking about getting a PHP whiz to help us out on this, but other things have been coming up. I’ll talk to him about it as soon as I get the chance.
I’m hoping that not only will it be permanent, but that a lot more can be done. If the theme can be reworked to drop/combine 10 or more separate files, and I can further improve the application processing time, I don’t see why the page loading time can’t be cut in half from where it is today. It probably won’t be done over a day or two like the changes I just made, but that goal should be doable without having to use a CDN.
One final update…
- I used pngcrush to make a major reduction on most of the PNG images.
- I changed the references to https://static.opensuse.org to be protocol relative, so users visiting the site over a non-secure connection don’t have to wait on SSL negotiation.
- Last but not least was fixing the “first byte” time. Turns out that we were having the forums software pull in some data from static.opensuse.org, which means that an internal request halfway across the world had to be completed before the site would even start responding to the client. I changed that code to use some caching, which cut ~1.5 seconds across the board.
Put that all together, and I think we’re just about there. Pages are now loading in half the time they were from last week.
Given the fact that we are working with off the shelf software, I don’t know if we’ll be able to easily use sprites or combine other files. However, mod_pagespeed may be a good way to get around some of that. I’ll probably look at that after the traffic from 12.2 settles down.
MatthewEhle wrote:
> Put that all together, and I think we’re just about there
I’d say it has improved. But I’m having to click buttons more than one
to get away from the working symbol in chromium of firefox. For
example: After login (which it’self is still slow), If I click the
‘Settings’ (usercp) to see posts I’m working with, it starts churning
in to action but nothing happens until I click it again, Same with ‘New
Posts’ ans sometimes with opening a post, I’ll have to close it (as
it’s just churning away), re-open it and bang it’s there…
It’s not my network. Because if I visit other forums/sites, it’s
lightening fast.
Thanks for your efforts though Matt
Certainly is a vast improvement here!
Thanks
I’m wondering if your problem is actually coming from the other issue that I have been battling with forums. Chrome will eventually return an empty response error, and Firefox will spin pretty much forever. It happens to a relatively few number of people, but it’s annoying as anything when it does.
After a lot of troubleshooting, Jim and I found that it is a problem with the server completing the TCP handshake with the client. Basically, the client properly returns the ACK, but the server doesn’t recognize it and keeps pumping out SYN+ACKs. Now that we know what to look for, it’s very easy to confirm. Basically, if you can capture your packets while seeing the problem, you will see a whole bunch of ACKs and SYN+ACKs in a row. It’s always preceded by a TCP window resize, although I’m not sure if that’s the trigger or just another symptom of the issue.
It’s almost certainly something on the kernel level, as control is not usually handed to the application until the handshake is complete and data packets are received. I have been doing a lot of research on this, and I have not found anything that exactly describes my problem. Core networking is my weak point in Linux, so that doesn’t help. In fact, if there are any people with experience in Linux network tuning or development, I would love to hear their input.