Cloud system with directadmin

C

casius

New member
Joined
Apr 28, 2010
Messages
4
Hi guys,

I just do not know how to solve our problem in regards to our busy server and directadmin.
Here is the fool story:
We have virtual machine in one of Germans DC on their cloud. The system is with:
2 CPU Quad Core Intel CPU
8 GB of RAM
500GB Xfs partition (/home)
30 GB ext3 for system

The thing is that we have about 9500 users in the system and their have filled up in one month. The problem started from short connection refuses, so searching for many of options (DNS, kernel, etc.) how to solve this problem, but the problem still was in operation. After that we have decided to write to DA support where they have suggested that connection refuses could be if server is busy and users are constantly added to the system, so apache just get restarted on and on, so in this case they suggested to insert graceful=1 to directadmin.conf.
We have added this thing, but the problem was still active.
Please anybody got any suggestions how we can solve unstable apache work?

Also I know that directadmin is not designed for high user amounts, but if that causes the problem (user amount) then I guess it is really not designed for todays technologies like cloud modules to handle busy server. I hope that there is some kind of round way of this solution, but just have no ideas about it :) I would be really appreciated of your help thank you.
 
In my understanding, a user is an account (Domain) in the /home/ folder. If you are talking about email users, then ignore the following:

My suggestion would be to setup DA servers, and not host anymore than approximately 1000 users per server. I might get a few people here disagree, but I feel this would be the limit I would allow in considering the following:

* Apache
* Memory usage for hungry apps like Joomla
* Mysql
* DNS
* Exim, Dovecot and any other Mail-related app

Then you need to look at storage also.

To me, 1000 users per box is the absolute premium. That means in your scenario you should be on your 10th server.

Now, this is going to sound a little contradictory - but if you configured that server (with the specs you showed) with virtual server instances (Xen or VMware), you could get away "theoretically" with running this kind of level of users. The reason here is because if you setup a virtual machine, you are giving it limits in resources. Rogue applications which chew up memory and CPU usage - will do exactly that - chew it all up - regardless if you have 1GHz CPU 500Mb RAM or 2 x Quad Core 2.5GHz and 8GB RAM.

Aside from all my rambings, you need to look at "top" (in our command-line) and sort by CPU and RAM usage to find out what is loading your server.

Apache can be tweaked to high load servers, and the following may be of assistance in configuring apache for this:

http://httpd.apache.org/docs/2.0/misc/perf-tuning.html

and this could also be a good site for you to check out:
http://www.crucialp.com/resources/t...ze-tweak-high-traffic-servers-apache-load.php
 
Thank you Ranz for reply. Here is my top for the cloud, but that does not says much, even the load is normal:
Code: 
top - 17:57:34 up 1 day, 18:41,  5 users,  load average: 6.37, 6.12, 6.19
Tasks: 297 total,   3 running, 293 sleeping,   1 stopped,   0 zombie
Cpu(s): 16.2%us,  7.7%sy,  0.0%ni, 75.8%id,  0.3%wa,  0.0%hi,  0.0%si,  0.1%st
Mem:   8388608k total,  8365404k used,    23204k free,   159668k buffers
Swap:  8388600k total,       96k used,  8388504k free,  3210004k cached

  PID USER      PR  NI  VIRT  RES  SHR S %CPU %MEM    TIME+  COMMAND
15380 root      25   0  566m 307m 4544 R 93.3  3.8   0:30.05 httpd
 1992 mysql     15   0 1085m 450m 5940 S 62.1  5.5 287:25.00 mysqld
14425 apache    15   0  703m 385m 6136 S  8.3  4.7   0:01.15 httpd
13668 apache    16   0  648m 383m 3516 S  6.3  4.7   0:07.42 httpd
15436 apache    16   0  656m 390m 4440 R  3.3  4.8   0:00.11 httpd
15569 apache    15   0  649m 385m 3884 S  2.3  4.7   0:00.07 httpd
14447 apache    15   0  650m 385m 5616 S  2.0  4.7   0:00.28 httpd
14501 apache    15   0  661m 397m 6012 S  1.3  4.8   0:00.66 httpd
15570 apache    15   0  645m 380m 3472 S  1.0  4.7   0:00.03 httpd
 7240 root      18   0  644m 383m 6732 S  0.7  4.7   3:47.88 httpd
14523 apache    15   0  669m 404m 6008 S  0.7  4.9   0:00.93 httpd
14532 apache    16   0  662m 398m 6520 S  0.7  4.9   0:01.81 httpd
15215 apache    15   0  649m 385m 3916 S  0.7  4.7   0:00.12 httpd
32754 root      16   0 13260 1632  820 S  0.7  0.0   4:52.50 top
14429 apache    15   0  666m 403m 6812 S  0.3  4.9   0:01.59 httpd
14515 apache    15   0  650m 386m 5612 S  0.3  4.7   0:00.91 httpd
14870 liyan     15   0 14080 2000  712 S  0.3  0.0   0:00.72 proftpd
15218 apache    15   0  654m 391m 4548 S  0.3  4.8   0:00.49 httpd
15274 apache    15   0  647m 383m 4196 S  0.3  4.7   0:00.23 httpd
15280 root      15   0 12872 1328  816 R  0.3  0.0   0:00.24 top
15559 apache    15   0  649m 383m 4328 S  0.3  4.7   0:00.09 httpd
15572 apache    15   0  647m 383m 3908 S  0.3  4.7   0:00.11 httpd
15573 apache    15   0  646m 382m 3756 S  0.3  4.7   0:00.07 httpd
15577 apache    19   0  644m 377m  756 S  0.3  4.6   0:00.01 httpd

The thing is that we have the major problem not with high traffic etc. even you have send me the link with 1.3 Apache version, but anyway the main problem is xfs filesystem that diretcadmin datasq works for about from 10-12 hours, the suspend and the termination process also are like a pain in the a**.
So I guess we should look to the problem with the actual xfs.
We also consider to get another partition of ext3 to re-loacate the user data files to get apache restarts faster, but that is just a solution for apache graceful restarts, but not the directadmin itself. Also we have removed the barriers for xfs, but that was unhelpful. In general we have about 9MB/S read speed which is really crap (even after optimization, before it was 5MB/s). Any thoughts?
 
The creator of ReiserFS (he's in prison now, for murder) would probably disagree with you. Years ago when Hans and I discussed it he told me that while it was very fast for reading very small files (and perhaps arguably good for public_html and Maildir) it's not fastest for writing, nor for large files (for example log files).

He told me to wait for Reiser4 before swtiching from ext3. I'm now using ext4 on new servers, and Reiser4 development appears to have stopped.

Jeff
 
That would be bad idea to change the filesystem in working server. Anyway we have noticed that our XFS partition and EXT3 partition are working on very low read speed (about 10MB/s). I guess the biggest issue that data storage switches are overloaded...We are getting the second machine and will do the same configuration, so we will see the difference.
 
You must log in or register to reply here.
Back
Top