one of the website takes a very long time to load
Hi,
I am using KVM and created four guest Operating systems on it. The server host is Ubuntu 10.04.I am using 4 websites in a reverse proxy environment. One of our website is running on CentOS VM. Right now there is no traffic on the website static HTML pages. I do not have any clue as why it was taking longer time to be accessed. Any suggestions as what might have happened ? There was a problem when I opened this thread and that problem got mixed in this thread if some one wants to read directly please check reply number 16 the actual problem is from there. http://www.linuxquestions.org/questi...ml#post3988551 you can skip in between. |
Please remember that when you write your OP that we know absolutely nothing. Sure we could read back all of your previous threads but that's rather inefficient. So please fill in the technical details:
- do all 4 websites run their own reverse proxy? - What kind of web stack components (software, versions, configuration) do these 4 websites run? - "Longer" in comparison to what? - What statistics go with that to prove it takes "longer"? - What website web stack daemon log excerpts prove it takes "longer"? - If the behaviour was OK before what software or configuration changes have happened? |
If there's no traffic on it, it won't be in the cache. But give a few statistics to show why you are worried.
|
Quote:
The configuration is like this Code:
|--------------192.168.1.1 Quote:
Quote:
Quote:
Quote:
Quote:
logs or some thing else. Quote:
eduCommons-3.2.1-1.x86_64.rpm on CentOS 5.5 in above diagram it refers to site5. Basically this eduCommons should be visible to world outside. Yesterday it was loading today morning when I opened this thread I came to know that it was taking longer time and by the time I am posting this message I noticed it (site5.abc.com) no longer runs. The eduCommon suite has probably installed Plone. I am new to Plone Zope. I looked in /var/log/httpd/ I see a file it says site5.abc.com-error_log Code:
[Thu May 27 11:23:20 2010] [error] (111)Connection refused: proxy: Code:
I see this and said it is not starting the Plone webserver. I found in /opt/eduCommons-3.2.1/var/log/ some logs I have put them here http://pastebin.com/KAe6dMVL and here http://pastebin.com/uGdLA82g I have no clue as what went wrong I am just guessing. |
It's not that I don't appreciate your verbose reply but looking at your log excerpts, wouldn't you agree that fixing your proxy errors would be the first thing to try?
|
Actually the time I posted the thread and by now the problem completely changed my site is right now not at all accessible on even LAN.
So the situation right now is even worse than for which I posted the problem. I will tell I am using a software eduCommons-3.2.1-1.x86_64.rpm on CentOS 5.5 which I am referring in the above diagram as site5.abc.com which is nothing but when some one hits in browser http://site5.abc.com reaches to my server A which in backend is redirecting to 192.168.1.5:8080/eduCommons on LAN before I posted the problem of slow access of the website I could browse http://192.168.1.5:8080/eduCommons we were designing templates and GUI for the site.So no where we messed with any settings internally. Code:
/etc/init.d/educommons restart This page http://educommons.com/documentation/...hat-5-x-centos says nothing much about installation which perfectly went fine and I was able to browse the site on LAN by Code:
192.168.1.5:8080/eduCommons In morning we thought of starting our work one of my team mates reported that site is loading slow. By the time I posted this message the site was very slow.We were not clear of the problem. But after that I rebooted server.Now nothing is there even on LAN.So the original problem has actually changed. I came across /opt/eduCommons-3.2.1/var/log where /opt/eduCommons-3.2.1 is the installation directory. Some one told me the error from Apache is because it can't access the eduCommons server. I tried killing all eduCommons processes and starting it again did not helped much. Some one suggested me that it is running a plone server and eduCommons is not able to connect to it. I do not know whether it is right or wrong. I googled and came across Zope.I am not able to confirm from my search if eduCommons is using that. This is one more file zeorserver.log I noticed things not working from 27th May but it might have gone wrong from 26th itself so see if in this file which is zeoserver.log I am not able to get which line should I focus on to be able to understand if some thing is wrong.I installed it on 25th May and it was working fine at that time Code:
2009-10-20T20:09:33 daemonizing the process |
Hi got my problem solved partially here by help of a friend of mine.
I will tell here is what he discovered but the original part from where discussion started is still there please read my next post. This post I have done so that if some one by chance comes across and faces similar thing should help them. There was a stale PID file that the educommons script seems to keep complaining about. He removed that file and Zope/Plone started up just fine. This is what he got when he started educommons: Code:
[root@o run]# /etc/init.d/educommons start And removed it. Then he started the services and it worked just fine: Code:
root@ var]# /etc/init.d/educommons start This is what he told me about stale pid " Most startup scripts on Linux use PID files for two purpose: 1) to keep track of services already running, and 2) to get the pid of the running service. Whenever we use an initscript to start a service, it will generally check for the existence of a PID file (as an indication of whether a service is running or not). When you shut down a service cleanly (ie. the service is stopped using an init script and does not crash or is not killed), the PID file is removed. So in cases when you DO NOT stop a service using an init script, the PID files get left over and become "stale". When the init script reported that "Client1" was already running, I checked the process list to see if the pid reported existed. It did not. So I gathered that this is a simple case of a stale PID file. The next few steps were about finding the PID file and removing it. ... after which the service started happily. As you can see, this can be a MAJOR irritant on any GNU/Linux system. There is service management system which he uses called 'runit' which "supervises" services. 'runit' is inspired by 'daemontools' (which DJB wrote for qmail). With 'runit' each service is supervised and a service superviser runs for each service. So we don't need to depend upon PID files to see if a service is running or not. Check out: http://smarden.org/runit/ for more information on this. " |
I have shared above message if some one by chance comes across this thread should help him.This problem occured in between now coming back to original problem of site loading slow.
I just used chromium to analyse why its taking so much time. (Not sure though) The problem has something to do with incorrect paths for the CSS & images inside the eduCommons web server. Though the page loads up very fast, the images don't load up at all. Here is a snapshot http://farm5.static.flickr.com/4071/...229bc3cb_o.png I have checked the web page from Google Developer tools and found some CSS missing.I myself have not developed eduCommons so can not understand how to analyse as what is making site slow.Assuming that CSS can not be a hurdle for site to be slow. So any guesses now? |
Quote:
What path do they reside in on the server? Were paths for CSS and images changed inside the eduCommons web server? Or did you by any chance change the eduCommons skin or other administration panel defaults? And there's no chance this has to do with your rewrite rules in the other thread, right? |
Quote:
No one knows about the site so no user also. eduCommons is hosted on CentOS Virtual Machine running as a Guest OS on Ubuntu 10.04 KVM hypervisor. To this server I have one more observation if I do a SSH from LAN it takes more than longer in comparison to other KVM hosts running on the top of same Ubuntu server. The networking is bridged networking. All this is a part of a LAN.All the KVM hosts can be pinged and SSHed but this (CentOS virtual machine) on which I am telling this site (eduCommons) is hosted is accessible on SSH also after a delay which is noticeable. All of this is running as Virtual Machines on a KVM hypervisor on Ubuntu 10.04 Quote:
|
Quote:
Chrome developer tools http://www.youtube.com/watch?v=RhaWY...ayer_embedded#! http://www.youtube.com/watch?v=OxW1d...layer_embedded read one more blog about what makes a website slow http://www.snaphow.com/why-is-my-web...e-blog-faster/ I recorded the website loading time in the Chrome time line section and based on what I observed following page elements take more than 100 ms on the website jquery checks 113 ms http://192.168.1.5:8080/eduCommons/p...achekey0517.js 244ms http://192.168.1.5:8080/eduCommons/p...achekey1443.js 247ms http://192.168.1.5:8080/eduCommons/p...chekey5402.css like this there are some more css and page elements which have taken 100ms or 200ms in their GET request from server. on this link http://www.chicagostyleseo.com/2010/...-site-is-slow/ they mentioned what Google has defined as a slow website following words are mentioned. "Google has made it clear that any process that takes more than 100 milliseconds (1/10 of a second) is too long" So that means my website is slow on the definition that Google gave above. I have attached a snapshot http://farm5.static.flickr.com/4048/...87025551_b.jpg you can see results of page elements loading slow. Also I checked LAN speed on Server A on above diagram. ethtool eth2 | less Code:
Settings for eth2: Code:
ethtool eth2 | grep -i speed Code:
ethtool eth0 ethtool eth0 | grep -i speed did not gave any output you can see above it is clear. Now what should I check? Or if any more statistics are needed? |
I take your point. 244ms is slow.
Accessing a site as 192.168.1.5(an internal ip) can have it's own issues and you should try and find which part of the access is causing a delay. Dns is out of it. If you go to the box in question, and type http://127.0.0.1/website http://localhost/website Do you still have the same delays? Does traceroute 192.168.1.5 throw up anything? |
Quote:
This step is one step further to resolve this problem. I was trying to run this from my local machine: Code:
ssh -L8080:localhost:8080 192.168.1.5 Quote:
The snapshot for this is here. To be able to try ssh -X 192.168.1.5 and then launch firefox/chrome and connect to http://localhost:8080/eduCommons so that I connect directly to eduCommons webserver which is running on Plone and bypass apache altogether. I was trying to install Chrome on CentOS VM (So that ssh educommons -X ;I can have a display of chrome running on CentOS server forwarded on my laptop ) by a method mentioned here http://www.if-not-true-then-false.co...-red-hat-rhel/ after getting an error like this Code:
http://dl.google.com/linux/rpm/stable/x86_64/repodata/repomd.xml: http://forums.fedoraforum.org/archiv.../t-207223.html which suggest to do Code:
yum clean all As you can see in bottom right hand corner of snapshot the speed with which CentOS VM is connecting to internet is 2.7 Kbps. Can this be a cause of my problem where as rest of the KVM hosts are having good internet speed. Another snapshot is here which shows a 37 Kbps speed for internet connectivity for a Ubuntu host. Notice the third last line of second snapshot which says Code:
Fetched 340kB in 9s (34.8kB/s) But have given snapshots for network connectivity. I could not understand the traceroute what exactly do you want me to test or execute in traceroute? |
This step is one step further to resolve this problem.
I was trying to run this from my local machine: Code:
ssh -L8080:localhost:8080 192.168.1.5 I got the same slow response 411 ms for an html element. The snapshot for this is here. To be able to try ssh -X 192.168.1.5 and then launch firefox/chrome and connect to http://localhost:8080/eduCommons so that I connect directly to eduCommons webserver which is running on Plone and bypass apache altogether. I was trying to install Chrome on CentOS VM (So that ssh educommons -X ;I can have a display of chrome running on CentOS server forwarded on my laptop ) by a method mentioned here http://www.if-not-true-then-false.co...-red-hat-rhel/ after getting an error like this Code:
http://dl.google.com/linux/rpm/stable/x86_64/repodata/repomd.xml: http://forums.fedoraforum.org/archiv.../t-207223.html which suggest to do Code:
yum clean all As you can see in bottom right hand corner of snapshot the speed with which CentOS VM is connecting to internet is 2.7 Kbps. Can this be a cause of my problem where as rest of the KVM hosts are having good internet speed. Another snapshot is here which shows a 37 Kbps speed for internet connectivity for a Ubuntu host. Notice the third last line of second snapshot which says Code:
Fetched 340kB in 9s (34.8kB/s) |
Ok after some failed attempts to install Chrome on CentOS here is what I found
on a centOS forum that Google does not support Chrome as in most of the cases CentOS or RHEL runs on a headless server which does not have GUI Here is a link https://www.centos.org/modules/newbb...topic_id=23746 this is where they discussed it originally http://www.google.com/support/forum/...5896c968&hl=en http://code.google.com/p/chromium/is...etail?id=26809 I was trying to install CentOS on the said server so that Code:
ssh -X root@192.168.1.5 Those who have been reading this thread if have some thing to try please let me know. |
All times are GMT -5. The time now is 10:49 PM. |