Server Performance... - PeachParts Mercedes-Benz Forum

Wayne 962 · #1 09-18-2011, 06:45 PM

Hi all. We moved the forums over to our new server farm last night. I'm not sure, but I think some of you may be having problems viewing the forums site? The new location will be faster than the old one when we get the configuration completed. On the old machine, the web server, the database, and the site search engine were all running on one machine. In the new config, they will be all separate so they can be optimized and can run faster. However, getting there sometimes involves a few blips in the process.

So, are the forums working okay for everyone here? Please let me know, so that I can double / triple check our new configuration.

Many thanks,

Wayne

Dubyagee · #2 09-18-2011, 06:53 PM

Hangs up every now and then when I hit New posts. Last time was an hour ago. Seems Ok now.

Bill Wood · #3 09-18-2011, 07:06 PM

For the most part, things are working well. The speed on the forums and the wiki is fine.

I am seeing what appears to be the MySQL server dropping out intermittently. This behavior is usually memory related. vBulletin is a memory hog and needs good PHP caching to perform well.

Things to look at:
vBulletin config.php
php.ini
my.ini

pawoSD · #4 09-18-2011, 10:29 PM

Lets hear the specs on this "farm"

And, why are you hosting from IIS?

Sigh.

Brian Carlton · #5 09-18-2011, 11:35 PM

Since the change, Firefox hangs nearly 50% of the time.

Chrome is a bit better but certainly not what it was before the change.

Don't be deluded into believing that it is working anywhere near correctly.

Aquaticedge · #6 09-18-2011, 11:45 PM

good ol PHP and mySQL, Both dedicated Pains in the ass.... Drove me mad to work with them.... Still miss them though.

Wayne 962 · #7 09-19-2011, 02:35 AM

Quote:

Originally Posted by Brian Carlton

Since the change, Firefox hangs nearly 50% of the time.

I have tested it on multiple computers with multiple browsers, and I can't seem to replicate any hanging issues? Firefox, iphone, Chrome, IE, Opera, Safari, etc. Perhaps if you can give me some specifics on your particular system that would help?

This also might be related to the DNS records populating across the Internet as well.

-Wayne

Stretch · #8 09-19-2011, 03:10 AM

At this time of the day (see post time stamp) things are usually really quiet. I rarely get any "trouble" until the back up kicks in at about 10:30 Dutch time! If I experience any problems during these quieter times I'll let you know - I'm guessing information from these times will help weed out problems related to being just too busy...

Wayne 962 · #9 09-19-2011, 05:35 AM

I've identified a problem that may be causing issues - it has to do with a limit on the number of connections to the database. For some reason, they are getting maxed out when they are queued. I don't seem to have this problem with the other server, I'll have to check on this setting.

I also just moved the search functionality to our dedicated server. This should significantly speed up daily operation *and* searching too.

-Wayne

Brian Carlton · #10 09-19-2011, 09:59 AM

Quote:

Originally Posted by Wayne at Pelican Parts

I have tested it on multiple computers with multiple browsers, and I can't seem to replicate any hanging issues? Firefox, iphone, Chrome, IE, Opera, Safari, etc. Perhaps if you can give me some specifics on your particular system that would help?

This also might be related to the DNS records populating across the Internet as well.

-Wayne

I've tried it on two different machines, at two different locations, both running Windows XP. Both have sufficient RAM (2GB).

When it hangs trying to load M-Shop, the other tabs for other sites will function just fine.

The issue is intermittent. This morning both Firefox and Chrome were both fine.

Brian Carlton · #11 09-19-2011, 10:43 AM

10:35 EDT:

It just took five minutes to get this page to load in Chrome.

Firefox is still trying................

You've got some serious issues..............

Brian Carlton · #12 09-19-2011, 10:44 AM

10:44 EDT:

Problem disappears.

Wayne 962 · #13 09-19-2011, 03:34 PM

Very odd. I have two machines here running the database as a master / slave. I have logging performed on the machines every one second (takes a snapshot of what they are working on)

Here's machine one at 10:42:58AM that says there's no pending requests on the database:

Code:

MySQL on localhost (5.0.51b-community-nt-log)                                                        up 0+06:56:46 [10:42:58]
 Queries: 5.7M   qps:  238 Slow:     0.0         Se/In/Up/De(%):    36/00/00/00 
             qps now:  356 Slow qps: 0.0  Threads:   13 (   7/  19) 38/00/00/00 
 Key Efficiency: 99.2%  Bps in/out:  38.8/ 7.2k   Now in/out:  41.3/ 7.4k

      Id      User         Host/IP         DB      Time    Cmd Query or State
--      ----         -------         --      ----    --- ----------
9577 mysql-log  localhost:5809      mysql         0  Query show full processlist
531221 mysql-log   FORUMS3:46336 vbulletin2         0  Sleep
531222 mysql-log   FORUMS3:46338 vbulletin2         0  Sleep
531226 mysql-log   FORUMS3:46342 vbulletin2         0  Sleep
531227 mysql-log   FORUMS3:46344 vbulletin2         0  Sleep
531228 mysql-log   FORUMS3:46346 vbulletin2         0  Sleep
531231 mysql-log   FORUMS3:46349 vbulletin2         0  Sleep
531232 mysql-log   FORUMS3:46351 vbulletin2         0  Sleep
531234 mysql-log   FORUMS4:4369 peachforum        0  Sleep
531236 mysql-log   FORUMS3:46354 vbulletin2         0  Query SELECT * FROM datastore WHERE title IN ('','options','bitfield
18221 mysql-log FORUMS-SEARCH:1
22776 Binlog Has sent all binlog to slave; waiting for binlog to be updated

Here's the 2nd machine (the search processor) that shows no queries pending for Peach either?

Code:

-----------------------------
MySQL on localhost (5.0.51b-community-nt-log)                                                   up 0+06:19:42 [10:42:58]
 Queries: 4.6M   qps:  212 Slow:     0.0         Se/In/Up/De(%):    36/00/00/00 
             qps now:   68 Slow qps: 0.0  Threads:   14 (   1/  13) 28/00/00/00 
 Key Efficiency: 99.6%  Bps in/out:  41.1/ 7.8k   Now in/out:  41.4/ 7.3k

      Id      User         Host/IP         DB      Time    Cmd Query or State                                                 --      ----         -------         --      ----    --- ----------                                                    327 mysql-log  localhost:1694      mysql         0  Query show full processlist                                      398494 mysql-log    FORUMS4:4500 peachforum         0  Sleep                                                                20      root  localhost:1693 vbulletin2         2  Sleep                                                            398421 mysql-log   FORUMS3:46337 vbulletin2        10  Sleep
398422 mysql-log   FORUMS3:46340 vbulletin2        10  Sleep
398426 mysql-log   FORUMS3:46345 vbulletin2        10  Sleep
398427 mysql-log   FORUMS3:46347 vbulletin2        10  Sleep
398429 mysql-log   FORUMS3:46352 vbulletin2        10  Sleep
398430 mysql-log   FORUMS3:46353 vbulletin2        10  Sleep
398433 mysql-log   FORUMS3:46355 vbulletin2        10  Sleep
398434 mysql-log   FORUMS3:46357 vbulletin2        10  Sleep
398436 mysql-log   FORUMS3:46359 vbulletin2        10  Sleep
398437 mysql-log   FORUMS3:46361 vbulletin2        10  Sleep
19      root  localhost:1692 vbulletin2     21051  Sleep 1 system us                                
22782 Connec Waiting for master to send event  2 system us                                  -10 Connec Has read all relay log; waiting for the slave I/O thread

So, it would appear to not be a database issue at this point. If it were a database slowness / pending issue, then it would show up as pending requests (seen this hundreds of times in the past when I've been tweaking the Pelican forums).

The only major change I have made (other than switching servers), was that I turned on gzip compression for the webpages. The previous server had this turned off, which causes the full download of all of the page code. Compressing it reduces the transfer time significantly.

I'll keep looking at the configuration to see what I can come up with.

-Wayne

Brian Carlton · #14 09-19-2011, 03:55 PM

Quote:

Originally Posted by Wayne at Pelican Parts

Very odd. I have two machines here running the database as a master / slave. I have logging performed on the machines every one second (takes a snapshot of what they are working on)

I'll keep looking at the configuration to see what I can come up with.

-Wayne

Thanks for the info.

If it were not for the fact that two different machines from two different ISP's exhibited the same issue on two separate days, I'd also conclude that it was a PC problem.

But, on each of the events, the PC would load other sites just fine.

I will continue to post when the issue rears its ugly head..............

Wayne 962 · #15 09-19-2011, 04:46 PM

Thanks, please let me know. It took me many years to get the Pelican sites up and running well. They get about 3X as much traffic as Peach, so we should be able to make this work. The #1 solution for Pelican was just adding more hardware (it runs on 3 servers). That, and a little bit of tweaking to the vBulletin code (which I haven't done on Peach yet), seemed to work.

I have a strange feeling the problem lies somewhere with the main server, not the database. I will take a closer look...

-Wayne