How many domains are you running?

[ispcomm]

This thread is old, but I would like to contribute.

I run an average of 150-250 domains per server using vhcs2 (modified for the purpose ;-). I also have an old server which is serving the same amount of domains (~200) with no control panel (hand installed and tuned). This old server is a dual P-III with 1024MB of ram and 3-disk soft raid-5 running on 120gbyte pata disks. The point is: any server should be enogh provided it's properly installed and tuned. I have notice of P4-3000 with 2Gbyte memory and hardware raid10 pushing an excess of 500 domains per server.

However, vhcs2 and omega (judging from some source code reading) will have trouble crossing the 100domain mark in their default configuration.

The problem is not how fast are the servers but the design and handling of logs, backups, statistics analysis etc. The continuous copying of files, analyzing and updating is killing performance as most of the time is spent copying and handling old data rather than serving sites.

The original vhcs2 I was using (3 years ago) hit it's dead point as soon as I reached the 50 domains mark. The problem gets even worse if a few of the sites are medium-traffic ones (I mean in the range of 1gbyte logs per day).

I would like to start a new thread for optimizing all this stuff (and post my mods) however I'm waiting for 1.0.0 to get out of the gates as any "distraction" would probably delay it's release.

ispcomm

[xister]

The big problem are the log files.
We have five open logs files for one domain.

see here:
http://www.isp-control.net/forum/max-dom...-2929.html
http://www.isp-control.net/forum/real-ti...-3137.html
http://www.isp-control.net/forum/log-cen...-2197.html

[rsmithgs]

ispcomm Wrote:This thread is old, but I would like to contribute.

I run an average of 150-250 domains per server using vhcs2 (modified for the purpose ;-). I also have an old server which is serving the same amount of domains (~200) with no control panel (hand installed and tuned). This old server is a dual P-III with 1024MB of ram and 3-disk soft raid-5 running on 120gbyte pata disks. The point is: any server should be enogh provided it's properly installed and tuned. I have notice of P4-3000 with 2Gbyte memory and hardware raid10 pushing an excess of 500 domains per server.

However, vhcs2 and omega (judging from some source code reading) will have trouble crossing the 100domain mark in their default configuration.

The problem is not how fast are the servers but the design and handling of logs, backups, statistics analysis etc. The continuous copying of files, analyzing and updating is killing performance as most of the time is spent copying and handling old data rather than serving sites.

The original vhcs2 I was using (3 years ago) hit it's dead point as soon as I reached the 50 domains mark. The problem gets even worse if a few of the sites are medium-traffic ones (I mean in the range of 1gbyte logs per day).

I would like to start a new thread for optimizing all this stuff (and post my mods) however I'm waiting for 1.0.0 to get out of the gates as any "distraction" would probably delay it's release.

ispcomm

Maybe it's worth starting this thread now? Wouldn't have thought there would be that many effecting factors between now and 1.0.0 stable?

[ispcomm]

xister Wrote:The big problem are the log files.
We have five open logs files for one domain.

Yes, their number and their handling is what I'm painfully aware of. Thanks for the pointers. I had already found the mod_log_spread one, but since my language is not german I hadn't found the other one.

I have studied the way gplhost solved their stuff (and for that matter I have studied all other open and many proprietary control panels released in the last 10 years). I have a solution for my own vhcs servers (but it breaks compatibility so it's a post 1.0.0 thing).

If you're curious, my solution only uses a single combined log file for all domains (and an extra error file, also for all domains). Logs are then tailed and fed in the database by a single process. A separate process filters them and feeds awstats. There's a house-keeping process that makes sure databases are compact and fast (i.e. not rrd as that looses too much data, but also well under the 2-3 million rows that vhcs would keep if I let it).

I could not believe it when I first discovered vhcs had set up a separate log file just for the bytes exchange on each request (that is just plain lazy and stupid).

ispcomm.

PS: Omega started when I was to fork vhcs by myself. So I've been lurking ispcp since the beginning and I've been hoping it would move on from where vhcs left. While seeing much effort from the developers, I won't hide my disappointment versus the actual results. Ispcp looks more like a facelift to vhcs instead of dealing with the structural problems that are present in it. This is why I decided to become more active in the development trying to help out amaic (as much as I can).

[ergo]

i have 30 domains on my VPS so far.

[KeNt]

I have IQ - 25 cm =)))

[momo]

I have 47 domaine and it is very smooth

[peman]

We have 5 so far.

[nuno]

so far 22 good product

[xister]

279 Domains and 82 Subdomains but a big problems with the logfiles (the old story).