Really poor performance compared to VB...

[mysqld]
socket=/var/lib/mysql/mysql.sock
pid-file=/var/run/mysqld/mysqld.pid
old-passwords
log-error=/var/log/mysql/error.log
log-slow-queries=/var/log/mysql/slowquery.log
long_query_time=5
connect_timeout=10
max_connections=1000
max_user_connections=1000
max_connect_errors=20
max_allowed_packet=2048M
key_buffer=32M
myisam_sort_buffer_size=64M
join_buffer=24M
record_buffer=32M
sort_buffer=2M
table_cache=1024
thread_cache_size=256
wait_timeout=120
tmp_table_size=32M
query_cache_size=64M
query_cache_limit=4M
innodb_buffer_pool_size=400M
innodb_file_per_table=1
innodb_additional_mem_pool_size=16M
[safe_mysqld]
open_files_limit=8192

[mysqldump]
socket=/var/lib/mysql/mysql.sock
max_allowed_packet=16M
add-drop-table
extended-insert
quick

[myisamchk]
key_buffer=32M
sort_buffer=32M
read_buffer=16M
write_buffer=16M

[mysql]
socket=/var/lib/mysql/mysql.sock
disable-auto-rehash
connect_timeout=10
local-infile
quick
#unbuffered
#silent

[client]
port=3306
socket=/var/lib/mysql/mysql.sock
 
Okay, IIRC you have 4GB RAM in your machine (and I'm going to presume you run other sites/DBs on the same server) so make a backup of your current config then try this one and see if it improves things:

Rich (BB code): 
[mysqld]
socket=/var/lib/mysql/mysql.sock
pid-file=/var/run/mysqld/mysqld.pid
old-passwords
log-error=/var/log/mysql/error.log
log-slow-queries=/var/log/mysql/slowquery.log
long_query_time=5
connect_timeout=10
max_connections=300
max_user_connections=1000
max_connect_errors=20
max_allowed_packet=8M
key_buffer=32M
myisam_sort_buffer_size=64M
join_buffer=512K
read_rnd_buffer_size=512K
record_buffer=32M
sort_buffer=512K
table_cache=4000
thread_cache_size=40
wait_timeout=120
tmp_table_size=64M
query_cache_size=32M
query_cache_limit=8M
query_cache_type=1
concurrent_insert=2
 
innodb_buffer_pool_size=1536M
innodb_file_per_table=1
innodb_additional_mem_pool_size=16M
innodb_flush_method=O_DIRECT
innodb_flush_log_at_trx_commit=2 
 
[safe_mysqld]
open_files_limit=8192
 
[mysqldump]
socket=/var/lib/mysql/mysql.sock
max_allowed_packet=16M
add-drop-table
extended-insert
quick
 
[myisamchk]
key_buffer=32M
sort_buffer=32M
read_buffer=16M
write_buffer=16M
 
[mysql]
socket=/var/lib/mysql/mysql.sock
disable-auto-rehash
connect_timeout=10
local-infile
quick
#unbuffered
#silent
 
[client]
port=3306
socket=/var/lib/mysql/mysql.sock

It might also help to increase the size of "innodb_log_file_size" - but that's more involved than simply making a change to my.cnf so we'll leave that one for now. ;)

Cheers,
Shaun :D
 
RobParker said:
I'm sure a lot of this is just getting the config set correctly but we're seeing really poor performance compared to what we managed with VB, particularly the number of users we handle online at a time.

With VB we could handle several thousand users online at once but after going live with XF a few days ago, when we get over 500 or so users we get errors that we've exceeded the max_user_onnections limit.

Is there anything obvious we should be looking to change in our config settings?
Click to expand...

I had 900 for a little while recently and it was running much faster than we ever did on vB4. We where slow for the first couple of days then found it was a modification that caused the problem.
 
There is clearly something wrong with your host setup. XenForo uses fewer resources than vBulletin.

As you maybe aware, coming from vBulletin; the more templates, modifications, skins, and other custom tweaks you add the more you add to the load.

Page Time: 0.2617s
Memory: 14.3025 MB (Peak: 15.4762 MB)
Queries (13, time: 0.0133s, 5.1%)

50+ modifications
10 skins
30 template edits
5 langues packs

150 members online

If your host claims you're using more resources and you're only using the stock (default, with nothing added) or less than me and some how you're numbers are higher than mine.... Clearly something is wrong with your setup.
 
RobParker said:
We're using hsphere and it's offloaded away from our server so control panel isn't an issue.
Click to expand...
Parallels is the company which makes Plesk and who also makes HS-phere. You may have found your issue. (maybe)

Checkout their support forum and amoung the top 5 (was #3 last I look) people complain about was innodb.
 
Just to update this, despite a huge amount of effort by our hosts we can't seem together XF running smoothly. They've concluded that XF is to blame and is just too inefficient for the resources we have. Obviously we don't believe this to be the case so we're planning on switching hosts to Nimbus Hosting who come highly recommended for busy XF installs.

An example of our problem is that we randomly see massive spikes in the number if MySQL threads that crash the server but aren't traffic related.
 
I looked through the thread and didn't notice what type of server it was - that is, how much CPU is dedicated to it, nor the size of the total DB........

Those server load figures are ridiculous! That said, forum which expect over 400-500 max users at once should be on a dedicated server (IMHO) or at the very least a very very capable VPS.

We have a single quad-core 2.6 MHZ - with XF and all the rest of our extensive site on it, plus some other smaller sites. We have not hit our busy time of year yet and don't have it optimized (no memcached or APC), but so far I have never seen CPU use exceed 20% nor seen a load of even 1.
This is with 350-400 users as reported by XF and my other CMS package....21,000 users and 1.2 million posts.

Something is very wrong in your case and in the case of a few others who have reported such loads! It stands to reason that if XF was really resource intensive...that spacebattles, IGN or Digital Point and other very large boards would have noticed it!

My old board (EE) used 20-50 db queries for each page - compared to 5-12 for most pages in XF.

I'll be able to report more once it gets colder (we are about heating!), but so far it seems that we will be easily able to handle our peak loads of 800+ users banging on it.
 
You must log in or register to reply here.
Top Bottom