Portal Home > Knowledgebase > Articles Database > Dedicated server very busy
Dedicated server very busy
|Posted by gfmileo, 04-14-2012, 09:50 PM|
I have a dedicated server 4GB ram, quad core AMD, Centos 6.2, 1GB/s dedicated speed with just apache installed in it to serve static mp4 files to a swf player. I've tried nginx with tons of different configs with php-fpm running as a socket for serving static files but it seems worst than apache, the load goes more than 200 when 50 or 60 users are online, with apache I can have 300 users downloading and it goes to 150 (I know it's very high).
This is a top command in a high traffic
When I did that top command the users were downloading at a rate of 326000 kb/s according to iptraf
I did an netstat and this is the result
My apache config is the default one, I haven't touch anything, I tried a bunch of different combinations but I had no luck.
Is there any tip or advice that you could give to me?
|Posted by net, 04-14-2012, 10:19 PM|
|Moved > Hosting Security and Technology.
|Posted by barbus, 04-15-2012, 03:55 AM|
|Most of apache process have status D. This may be due to the high load on disk system. Considering specificity work of your server, it is quite possible.
Show output. This will show full picture.
|Posted by gfmileo, 04-15-2012, 04:58 PM|
|I have no results with that code.
|Posted by ArturasLIX, 04-16-2012, 01:04 AM|
|Yes, the high load is due high IO system usage. Perhaps someone attacks your server by accessing some script that loads system. Check Apaches logs to see what's happening. For instance:
|Posted by barbus, 04-16-2012, 01:51 AM|
|In such a case you need to add additional repository RPMforge for Centos 6.X. In base CentOS repository dstat is missing.
and try install dstat
|Posted by ArturasLIX, 04-16-2012, 02:52 AM|
|dstat is in base CentOS repository, no need to download package, it's enough to just run:
But I can't see the reason why should he run dstat? From provided top output it's quite clear that load was generated due high IO operations:
In my opinion, we should help to find the reason of that.
|Posted by gfmileo, 04-16-2012, 07:24 PM|
|You're very right about that. But I will do anything that you ask me to just to fix that problem with the server.
|Posted by gfmileo, 04-16-2012, 07:28 PM|
|This is the dstat 10 result in a non peak hour...
|Posted by ArturasLIX, 04-17-2012, 04:04 AM|
|Do you know where your apache logs are? At the high load time I suggest you to tail that logs, to see the queries that loads the system, for instance:
After that you will see the queries and the IPs that makes the load.
|Posted by trige, 04-17-2012, 03:53 PM|
|in most cases there is a hard load on HDD or your hdd is being broken.
To find hdd load i suggest to install such a program like iostat.
To troubleshoot HDD use smartctl.
|Posted by gfmileo, 04-17-2012, 05:48 PM|
|This is the result for iostat
|Posted by ArturasLIX, 04-18-2012, 01:47 AM|
|What is the point of suggesting to ensure that it's a HDD load with many different ways? How can it help find the real reason of that? I think it's clear from the first post, that it's a HDD load, now you need to find the reason. Also from first post it's clear that it's because of web server`s work. To ensure, there is tool that shows IO usage by process:
But as I said, you should tail your websites logs to find the website or IP that abusing your system, then block it.
|Posted by ZenMonk, 04-18-2012, 03:27 AM|
|What kind of HDD setup do you have in there? RAID10? Also, go for 15K RPM drives if you really want to get your load down. Looks like you are on the low end 7.2K RPM drives.
|Posted by RRWH, 04-18-2012, 06:44 AM|
|IO is the issue - to get a better picture,
iostat -x -d 10 10
iostat -d 10 10
What is the disk subsystem in use? How many disks, software or hardware raid?
Find what is causing the high IO and address it - either with more disks or breaking up the IO workload across several servers.
Add to Favourites Print this Article