BigTex
Retired WS Staff
- Joined
- Feb 25, 2005
- Messages
- 118
- Reaction score
- 119
Hey guys - I thought i would start a new thread that gives some information on what is going on with our server and what is being done to address the issues. I don't want to get too technical but will a little so you can understand the complexity of running a webserver like WS. Let me start by pointing out that the server is the computer hardware and the webserver is software that runs on the server that handes serving web pages to clients (like your Internet Explorer or Firefox).
First off the various issues we have had with the hosting company are partially misconceptions on what a server hosting company does versus what a web hosting company does. A server hosting company (like ours) provides a server with cooling and power and security but very little application level support (as opposed to a webhosting company which does it all but is *considerably* more expensive). When WS stops working the hosting company will only help by rebooting the box (which helps but is much too severe of an action to be considered a reasonable fix - like replacing the light fixture every time a bulb burns out) and not fix the webserver application (which is what you actually use when you login). I have a server in the same datacenter as the WS server and have never had downtime nor support issues in the 4 years there but that is because I support the applications myself and only use the hosting company for hardware issues. Changing hosting companies is usually done for geographical or power/cooling stability issues (or simply cost).
Now on the hardware side I do see several issues that I am working on, all centered on the memory subsystem. First and foremost the server had no swap space allocated. Swap space is a temporary space on the disk where the server can move memory to when it is running out of real memory (RAM, which is on a chip). RAM is always better than swap (because it is an order of magnitude faster) but cost of RAM prohibits having an excessive amount and swap addresses when you need some slower temp memory when under load. The lack of swap space meant that when the webserver got enough users on it the server would run out of memory and get stuck and need a complete system reboot to fix it. I added swap space and this has resulted in no server hangs (remember server means hardware) but we are still seeing webserver hangs.
On the webserver issues I am looking at logs when it locks up and trying to identify the trigger mechanism that causes the hang. This means I might need WS to lockup several more times before we have enough data to do a trend analysis on the failures. Please bear with me while I work on it.
Finally on the signature removal and why it helps. All the data stored on the WS webserver is put in a back end mysql database. That means that all the information for each user,post,thread,etc is put into and read out of the database. Since the goal is to stabilize the server, the extra work caused by the signatures (especially the larger ones) and subscribed threads just add to the work the server needs to to every moment. If we can stabilize the server then we can re-evaluate signatures but even then I would strongly suggest not allowing the large pictures sigs.
I know many of you are not as interested in what is going on as you are in just seeing WS available so if you just skipped to this part please be aware that I am working hard to get this addressed and ask your patience while I complete my work.
Sincerely,
-BigTex (Sassygerl's Hubby)
ETA: Fixed the "we" when I meant me, I am used to writing corporate explanations with my entire team rather than just me.
First off the various issues we have had with the hosting company are partially misconceptions on what a server hosting company does versus what a web hosting company does. A server hosting company (like ours) provides a server with cooling and power and security but very little application level support (as opposed to a webhosting company which does it all but is *considerably* more expensive). When WS stops working the hosting company will only help by rebooting the box (which helps but is much too severe of an action to be considered a reasonable fix - like replacing the light fixture every time a bulb burns out) and not fix the webserver application (which is what you actually use when you login). I have a server in the same datacenter as the WS server and have never had downtime nor support issues in the 4 years there but that is because I support the applications myself and only use the hosting company for hardware issues. Changing hosting companies is usually done for geographical or power/cooling stability issues (or simply cost).
Now on the hardware side I do see several issues that I am working on, all centered on the memory subsystem. First and foremost the server had no swap space allocated. Swap space is a temporary space on the disk where the server can move memory to when it is running out of real memory (RAM, which is on a chip). RAM is always better than swap (because it is an order of magnitude faster) but cost of RAM prohibits having an excessive amount and swap addresses when you need some slower temp memory when under load. The lack of swap space meant that when the webserver got enough users on it the server would run out of memory and get stuck and need a complete system reboot to fix it. I added swap space and this has resulted in no server hangs (remember server means hardware) but we are still seeing webserver hangs.
On the webserver issues I am looking at logs when it locks up and trying to identify the trigger mechanism that causes the hang. This means I might need WS to lockup several more times before we have enough data to do a trend analysis on the failures. Please bear with me while I work on it.
Finally on the signature removal and why it helps. All the data stored on the WS webserver is put in a back end mysql database. That means that all the information for each user,post,thread,etc is put into and read out of the database. Since the goal is to stabilize the server, the extra work caused by the signatures (especially the larger ones) and subscribed threads just add to the work the server needs to to every moment. If we can stabilize the server then we can re-evaluate signatures but even then I would strongly suggest not allowing the large pictures sigs.
I know many of you are not as interested in what is going on as you are in just seeing WS available so if you just skipped to this part please be aware that I am working hard to get this addressed and ask your patience while I complete my work.
Sincerely,
-BigTex (Sassygerl's Hubby)
ETA: Fixed the "we" when I meant me, I am used to writing corporate explanations with my entire team rather than just me.