Server Down - My Idea about it.

[dextergo]

This is my understanding. Please correct me if I am wrong.

In a shared hosting, the hosting company puts as much customers in a server.

When he puts too much and the CPU or Hard Drive reaches maximum, the server will hang.

When the server hangs. Everyones website is down.

My question:

1.) Do the servers normally restart automatically after this happens? Or it has to be done manually by the hosting company.

2.) What does the hosting company do to fix this server? Move accounts to another server to free up space?

3.) Is there an automatic system that moves accounts to another server before the original server reaches maximum capacity?

4.) If the server is getting to maximum capacity, is it advisable to move clients to another server? If yes, will this cause downtime for the client?

5.) If the server is getting to maximum capacity, is it better to stop adding clients or just free up space by moving clients to another server?

 

[RH01]

Sometimes this is correct with overloading. Not always though.

for your answers

1. Servers only automatically restart only if they crashed entirely and are in reboot mode. Usually they have to be done with a reboot ticket to the data center so manually.

2. All depends on the problem. If it is truly overloading then they need to get another server and move the accounts over to balance it. If its another issue then they need to go in and find the reason and fix it, have the data center do it or hire someone to do it.

3. Depends on the softwaree running, Cpanel can automatically move accounts when told to do so and its pretty easy

4. 4. Already answered that in question 2

5. Depends on the problem and the clients. Could just be one problem client. Could be a reseller too

 

[mymoonlight]

1.) Do the servers normally restart automatically after this happens? Or it has to be done manually by the hosting company.
-->It's quite complex, most time, a auto restart will bring the server up ,but some time,if there is a attack, ,restart won't work.

 

[wschuah]

1) The server need to restart manually.
2) Depend of the cause. If data overstoring, it should be the way.
Beside that they can add in more hard disk if available..
3) Not in hardware side..
4) Sure.
5) depend your what you want, stop means no incoming. free up means you need invest..

 

[Zishan]

I would say that it truly depens upon the providers that how much accounts they put on a single server. Not necessarily that all servers will be of same specs so the number of accounts may vary depending upon the specifications of the server. Most of the companies do over sell but they have a strict limit of accounts per server. I would answer your questions as:

1) If the server load is high and the server is hanging then a manual restart is required. That can be done through a remote reboot switch ( e.g APC) or by contacting the DataCenter techs. Other possible way is to setup some load monitoring scripts that continue to monitor the server load and if the server load reaches a specified number, then the script initate a reboot just to bring the things back to normal.

2) If the company is just over selling and over utilizing the hardware and there are more number of domains than the hardware can handle then hosting provider should split the domains and move the accounts of the server to free up its resources so that the domains/users on that server can utilize the resources properly. Space doesn't matter always. Sometimes a single domain crashes the entire server and sometimes a server with more than 100 domains runs fine. It all depends that how much traffic or how much resources each domain requires. If any user/domain requires more resources than the shared server can offer, that user/domain should be advised an upgrade so that the other users will not be affected.

3) By automatic system if you mean that no human interaction then I don't think so. Dealing with user's data is a critical task so a human should be there to keep an eye on it. Different control panels offer different ways of packaging and restoring accounts so such methods are used to move the account from one server to another so that all user setting remain intact.

4) I would say that there should always be a number defined lower than the maximum capacity of the machine. This will give the benfit that the clients on that server will not have any issues. If by any chance, server gets to maximum capacity then the accounts transfer may cause negligible downtime. It depends that how the data transfer is being made. If account transfer involves IP change then it will involve downtime due to DNS propagation but if any account has dedicated IP then that IP can be moved to the new machine too and the account will be up and running from new server withine few minutes.

5) A server audit should always be there to identify that whether the server can handle more accounts or not. If its load condition is fine and it doesn't show any problem then few accounts can be added to it but if its already in bad condition or just doing well and can't handle more accounts then there shouldn't be a single account more on that server and if possible some heavy users should be moved off to stabilize it.

 

[BobZ]

1.) Do the servers normally restart automatically after this happens? Or it has to be done manually by the hosting company.

Using it is done either by means of the APC reboot system or manually by the datacenter support representatives.

 

[~ServerPoint~]

5.) If the server is getting to maximum capacity, is it better to stop adding clients or just free up space by moving clients to another server?

That really depends on the web hosting provider. Some of them have 50% as read line some more or less.

 

[SoftsysHosting]

In a shared hosting, the hosting company puts as much customers in a server.

This is not necessarily true. Non-oversold and quality hosting solutions always have limits setup on the number of websites that are hosted on a server. Also, the server load is monitored on a timely basis in order to ensure that the server is not over loaded. With this in mind, when the server resources reach a threshold (set by the host), a new server is setup for new orders/websites. This is the procedure that we follow and I must admit that it has helped us tremendously in maintaining high quality of services minimizing any incidents of server crashing, etc. I am sure that there are many companies around the world that are following a similar strategy.

- Rick

 

[~ServerPoint~]

Rick you are right. Even some companies share this information with their customers.

 

[ldcdc]

I am sure that there are many companies around the world that are following a similar strategy.

Even onverselling ones. Fact is, it makes very little sense to keep a server overloaded, unless you don't provide support either. In that case, the extra tickets complaining about server performance don't bother you.

 

[BobZ]

This is not necessarily true. Non-oversold and quality hosting solutions always have limits setup on the number of websites that are hosted on a server. Also, the server load is monitored on a timely basis in order to ensure that the server is not over loaded. With this in mind, when the server resources reach a threshold (set by the host), a new server is setup for new orders/websites. This is the procedure that we follow and I must admit that it has helped us tremendously in maintaining high quality of services minimizing any incidents of server crashing, etc. I am sure that there are many companies around the world that are following a similar strategy.

Some companies really do so, but most of them simply make fools of their clients saying that the server is up and running while it is overcrowded with thousands of clients who constantly overload the server and overuse all the resources.

Such companies will change their server to a new one only in case it gets crashed or some server partitions get completely broken.

 

[~ServerPoint~]

Such companies will change their server to a new one only in case it gets crashed or some server partitions get completely broken.

Poor customers.... But that is the way to lose all of them

 

[BobZ]

Sometimes the customers are simply offered a tranfer to a brand new server and this saves some of the clients for the hosting company.

 

[mikeyellob]

Here are my answers based upon my experience,

A1.) The situation you are explaining resulted in hanging of system and not crashing which doesn't hook auto reboot. Hence in this case, it will be manual reboot.

2.) Usually this is resolved after looking at server log as what was the reason of this. Usually websites requiring CPU extensive operations are moved to less populated server. In this way, moving a few websites resolved this issue.

3.) Well!... as per my knowledge there is none. However you can set up traces to diagnose such situations and then can take necessary precautions.

4.) Again it depends upon the situation. IN case of CPU Bottle Neck, yes we need to move more CPU Extensive applications to other server to resolve this.

5.) As per practice, every company has its own policy to host a MAX number of websites on one single server. They keep on moving CPU Extensive sites to other servers until unless they reach the MAX number.

Thanks,

Mike

 

[BobZ]

As per practice, every company has its own policy to host a MAX number of websites on one single server.

Yeah, that is right. Still there are companies that violate their AUP and host more accounts on a single server than it is allowed by their policy

 

[ldcdc]

Yeah, that is right. Still there are companies that violate their AUP and host more accounts on a single server than it is allowed by their policy

With the exception of semi-dedicated accounts, the vast majority of hosts don't have an openly stated policy regarding the number of accounts or websites they host per server. I suppose that an internal one, or a way to determine internally how many accounts they should place per server, does exist at most hosts, even if only as vague as "as soon as the load is >X for Y hours" or "as soon as the CPU usage is >W for more than Z hours".

 

[AbbieRose]

So how easy is it to move an account to a new server and what are the risks whilst doing so-is there a chance of data loss? Do they have to do this manually when they notice there being too high a server load, or is there an automatic, kind of floating system?