My name is Tim, and I work at Digital Pacific. I thought I might just come and debunk a few of the myths that were displayed on this forum.
Chanboy wrote:I checked out the Digital Pacific website and some of their other clients, all seem to be performing ok.
This is true. No other clients of ours were getting any serious performance losses apart from bicycles.net.au. At Digital Pacific, we monitor all sites that have more then 500MB worth of traffic a day, this is on top of per-server monitoring. With this site monitoring, we are aiming to identify sites that are getting high load(s), and if necessary move the site to a more powerful server.
This is the graph that was shown from our per-site monitoring for bicycles.net.au. To provide you all with a bit of history, the server before around 7:30PM (AEST / GMT+10) was moved already to three servers in different configurations in one day to find out what configurations were working or not.
Chanboy wrote:Software configuration - more likely what my guess would be.
Well, at this moment the team working on the performance issues on Wednesday are still not sure exactly where the issue was, but to narrow it down, it has something to do with the visualization technology "Virtuozzo". When the site was moved to any server that did not have Virtuozzo, the site started performing well.
Chanboy wrote:I would be investigating the setup of the server and also the forum software to see if any such configuration has been overlooked.
This is what Chris on Tuesday evening asked us to do, and for twelve hours on Wednesday, this is what four staff members (including me) were doing. On the phone while speaking to Chris, I told him that it could take anywhere between ten minutes to three days to find out where the bottleneck was. By 4pm that day, we had the website working on a server at a very acceptable speed. When we found that this configuration worked, we put the server in datacentre, and did some internal load testing. At 8pm, when we were sure that everything was working well, we decided to do the move to the new server.
Using the information provided Chanboy I have filed a job with Digital Pacific, job no. OKJ-584202. Don't know if it will help or not but at least they know they have upset users.
Andrew (and others), we already knew that customers of bicycles.net.au were being affected, and we were working as hard as possible to fix these issues. At the end of the day, it was a mix of different things (both our fault - in the case of Virtuozzo) and the fault of the software (such as Typo3) that were causing serious performance issues.
Rockford wrote:Looks like they have moved you to your own server. I can't find any other domains being hosted on that IP 126.96.36.199
But if the other domains still hosted on .81 are having issues it could be a Datacentre issue or a switch/router issue.
Ever since the upgrade, bicycles.net.au has been on a Virtuozzo virtual environment, and the only site hosted on the IP was bicycles.net.au. The issue was never related to any networking issues.
Since the move to a dedicated server without any Virtuozzo technology running in the background, the site will now be running smoothly. I have made a few changes to the Typo3 config to ensure that there are no issues with MySQL connections all being taken up, as what was happening previously.
The virtuozzo container that Bicycles had did not have any limits to memory, disk or CPU usage, and the new dedicated server has a very similar configuration. Memory utilization on the old container was never more then 50%, so memory as well was never an issue.
The technical team here at Digital Pacific will be working hard to find out why when the website is hosted in the Parallels Virtuozzo environment the site is slow to load, and file a bug report with Parallels, the company who provides the software.
I hope that you understand that we have been working hard for a whole week to ensure that response times for the website were improved. We knew about the delays. Digital Pacific had a team of four working Wednesday for twelve hours to fix the issues that everyone have been seeing for the past week.
Hope this helps you all understand the issues that were being faced,